Papers with optimization
Unveiling the Deficiencies of Pre-trained Text-and-Layout Models in Real-world Visually-rich Document Information Extraction (2026.findings-eacl)
Copied to clipboard
Chong Zhang, Yixi Zhao, Yulu Xie, Chenshu Yuan, Yi Tu, Ya Guo, Mingxu Chai, Ziyu Shen, Yue Zhang, Qi Zhang
| Challenge: | PTLMs have shown remarkable success in multiple information extraction tasks . however, their performance in real-world scenarios falls short of expectations . |
| Approach: | They propose to use an entity-centric dataset to evaluate PTLMs' performance . they find that inadequate annotations in benchmark datasets lead to spurious correlations . |
| Outcome: | The proposed dataset disentangles the falsely-coupled segment and entity annotations that arises from the block-level annotation of FUNSD. |
Knowledge-Enriched Natural Language Generation (2021.emnlp-tutorials)
Copied to clipboard
| Challenge: | Knowledge-enriched text generation poses unique challenges in modeling and learning . a roadmap will outline the state-of-the-art methods to tackle these challenges . |
| Approach: | They propose a roadmap to tackle the challenges of knowledge-enriched text generation . they will dive deep into various technical components to illustrate how to represent knowledge . |
| Outcome: | This tutorial outlines the state-of-the-art methods to tackle the problem . it aims to show how to represent knowledge, feed knowledge into a generation model, evaluate results . |
Intuitive Fine-Tuning: Towards Simplifying Alignment into a Single Process (2025.acl-long)
Copied to clipboard
| Challenge: | Supervised Fine-Tuning (SFT) and Preference Optimization (PO) are key processes for aligning Language Models with human preferences post pre-training. |
| Approach: | They propose to combine Supervised Fine-Tuning and Preference Optimization (PO) with two sub-processes defined at token level within the Markov Decision Process (MDP) |
| Outcome: | The proposed process performs comparably or even superiorly to SFT and some typical PO methods across several tasks, particularly those requires generation, reasoning, and fact-following abilities. |
Deep Bayesian Natural Language Processing (P19-4)
Copied to clipboard
| Challenge: | Introduction to deep Bayesian learning for natural language addresses the fundamentals of statistical models and neural networks. |
| Approach: | This tutorial addresses the advances in deep Bayesian learning for natural language . it focuses on advanced Bayessian models and deep models . authors present case studies and domain applications to tackle different issues . |
| Outcome: | This tutorial focuses on advanced Bayesian models and deep models for natural language . case studies and domain applications are presented to tackle different issues in deep Bayessian processing, learning and understanding. |
Guiding Variational Response Generator to Exploit Persona (2020.acl-main)
Copied to clipboard
Bowen Wu, MengYuan Li, Zongsheng Wang, Yifu Chen, Derek F. Wong, Qihang Feng, Junhong Huang, Baoxun Wang
| Challenge: | Neural Response Generators (NRGs) use persona information of users to perform personalized conversations . current studies focus on incorporating explicit meta-data of user profiles or character descriptions to generate persona-aware responses. |
| Approach: | They propose to use persona information of users in Neural Response Generators to perform personalized conversations. |
| Outcome: | The proposed method improves persona-aware response generation and the metrics are reasonable to evaluate them. |
Modeling the Q-Diversity in a Min-max Play Game for Robust Optimization (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for minimizing the worst-case loss of annotated groups are lacking in practice due to expensive annotations and privacy issues. |
| Approach: | They propose a distributionally robust optimization framework that relaxes group identification into direct parameterization by using an interactive training mode. |
| Outcome: | The proposed method outperforms state-of-the-art methods on synthetic and real-world text classification tasks. |
ORMind: A Cognitive-Inspired End-to-End Reasoning Framework for Operations Research (2025.acl-industry)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have shown promising results in various domains, but their practical application in industry-relevant operations research presents significant challenges and opportunities. |
| Approach: | They propose a cognitive-inspired framework that enhances optimization through counterfactual reasoning . they use a workflow that transforms requirements into mathematical models and executable solver code . |
| Outcome: | Experiments show that ORMind outperforms existing methods in the NL4Opt dataset and ComplexOR dataset. |
Long-Distance Dependencies Don’t Have to Be Long: Simplifying through Provably (Approximately) Optimal Permutations (P19-2)
Copied to clipboard
| Challenge: | Neural models at the sentence level often need to model the interaction between words . however, there is no guarantee that the standard ordering of words is computationally efficient or optimal . |
| Approach: | They propose to use a dependency parse as a proxy for inter-word dependencies in a sentence to simplify the sentence with combinatorial objectives imposed on the sentence-parse pair. |
| Outcome: | The proposed model improves classification accuracy and reduces classification error by 2.0% over the previous state of the art. |
Arxiv Copilot: A Self-Evolving and Efficient LLM System for Personalized Academic Assistance (2024.emnlp-demo)
Copied to clipboard
| Challenge: | Existing solutions for document QA fail to provide personalized and up-to-date information efficiently. |
| Approach: | They propose to deploy a self-evolving, efficient LLM system that can offer personalized research services, maintaining a real-time updated database. |
| Outcome: | The proposed system saves 69.92% of time after efficient deployment. |
Cyclical Annealing Schedule: A Simple Approach to Mitigating KL Vanishing (N19-1)
Copied to clipboard
| Challenge: | Variational autoencoders (VAEs) with an auto-regressive decoder have been applied for many natural language processing tasks. |
| Approach: | They propose a cyclical annealing schedule which repeats the process of increasing multiple times to learn more meaningful latent codes progressively by leveraging previous learning cycles as warm re-restart. |
| Outcome: | The proposed method improves on a broad range of NLP tasks, including language modeling, dialog response generation and semi-supervised text classification. |
UNSEE: Unsupervised Non-contrastive Sentence Embeddings (2024.eacl-long)
Copied to clipboard
| Challenge: | Unsupervised Non-Contrastive Sentence Embeddings demonstrates better performance compared to SimCSE in the Massive Text Embing (MTEB) benchmark. |
| Approach: | They introduce UNSEE, which stands for Unsupervised Non-Contrastive Sentence Embeddings, which demonstrates better performance compared to SimCSE in the Massive Text Embing benchmark. |
| Outcome: | The proposed solution achieves better performance than contrastive objectives on the Massive Text Embedding (MTEB) benchmark. |
Reciprocal Learning of Knowledge Retriever and Response Ranker for Knowledge-Grounded Conversations (2022.coling-1)
Copied to clipboard
| Challenge: | Recent work on grounding dialogue agents with knowledge documents has sparked increased attention . hand-labeling data to that end is time-consuming and many datasets lack knowledge annotations . |
| Approach: | They propose a reciprocal learning approach to optimize a knowledge retriever and a response ranker for knowledge-grounded response retrieval without ground-truth knowledge labels. |
| Outcome: | The proposed model outperforms previous state-of-the-art methods on two public benchmarks. |
When Prompt Optimization Becomes Jailbreaking: Adaptive Red-Teaming of Large Language Models (2026.eacl-srw)
Copied to clipboard
| Challenge: | Existing safety evaluations rely on fixed collections of harmful prompts . such attacks span single-shot prompts, multi-turn interactions, cross-lingual settings . |
| Approach: | They propose to use black-box prompt optimization techniques to search for safety failures . they use GPT-5.1 to optimize for a continuous danger score . |
| Outcome: | The proposed approach reduces effective safeguards for large language models . the average danger score of Qwen 3 8B increases from 0.09 in its baseline setting to 0.79 after optimization. |
Symbolic Prompt Program Search: A Structure-Aware Approach to Efficient Compile-Time Prompt Optimization (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent work on prompt programs has focused on simple prompt programs or assumed that the structure of a prompt program is fixed. |
| Approach: | They propose a framework to perform symbolic prompt program search for compile-time optimizations of prompt programs. |
| Outcome: | The proposed framework improves performance of complex prompts on instruction tuning, pipeline tuning, prompt compression and more. |
EvoAgentX: An Automated Framework for Evolving Agentic Workflows (2025.emnlp-demos)
Copied to clipboard
| Challenge: | Existing MAS frameworks often require manual workflow configuration and lack native support for dynamic evolution and performance optimization. |
| Approach: | They propose an open-source platform that automates generation, execution, and evolutionary optimization of multi-agent workflows. |
| Outcome: | The proposed platform automates generation, execution, and evolutionary optimization of multi-agent workflows. |
Structural Reward Model: Enhancing Interpretability, Efficiency, and Scalability in Reward Modeling (2025.emnlp-industry)
Copied to clipboard
Xiaoyu Liu, Di Liang, Hongyu Shan, Peiyang Liu, Yonghao Liu, Muling Wu, Yuntao Li, Xianjie Wu, Li Miao, Jiangrong Shen, Minlong Peng
| Challenge: | Generative RMs (GRMs) lack contextual and background information during inference, leading to incomplete evaluations. |
| Approach: | They propose a modular and interpretable framework that integrates side-branch models as auxiliary feature generators. |
| Outcome: | The proposed framework outperforms scalar and saline reward models in robustness and alignment with human preferences. |
Test-time Adaptation for Machine Translation Evaluation by Uncertainty Minimization (2023.acl-long)
Copied to clipboard
| Challenge: | evaluators of machine translation systems often use text-based metrics to evaluate performance . however, these metrics lack semantic-level information and exhibit poor correlation with human ratings . authors propose a method to reduce inference bias of neural metrics in out-of-distribution data . |
| Approach: | They propose to reduce inference bias by using uncertainty estimation, test-time adaptation, and inference to reduce model uncertainty. |
| Outcome: | The proposed method reduces model uncertainty and improves correlation performance across models. |
RatE: Relation-Adaptive Translating Embedding for Knowledge Graph Completion (2020.coling-main)
Copied to clipboard
| Challenge: | Existing approaches for knowledge graph embedding have limitations in complex vector space . embeddability of one-to-many relations is not explicitly alleviated . |
| Approach: | They propose a relation-adaptive translating embedding function that can be extended to complex vector space. |
| Outcome: | The proposed translation function improves expressive power and alleviates embedding ambiguity problem. |
MUSE: A Multimodal Conversational Recommendation Dataset with Scenario-Grounded User Profiles (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing research focuses solely on text, leaving a gap with practical applications. |
| Approach: | They propose to synthesize a multimodal conversational recommendation dataset using multimodal large language models to automatically synthesized data from 7,000 conversations in the Clothing domain. |
| Outcome: | The proposed dataset contains 83,148 utterances from 7,000 conversations centered around the Clothing domain. |
In-context Learning and Gradient Descent Revisited (2024.naacl-long)
Copied to clipboard
| Challenge: | In-context learning (ICL) has shown impressive results in few-shot learning tasks, yet its underlying mechanism remains elusive. |
| Approach: | They propose a simple gradient descent-based optimization procedure that respects layer causality and improves similarity scores significantly. |
| Outcome: | The proposed procedure improves similarity scores on untrained models despite not showing ICL. |
GRAM: Fast Fine-tuning of Pre-trained Language Models for Content-based Collaborative Filtering (2022.naacl-main)
Copied to clipboard
| Challenge: | Content-based collaborative filtering (CF) predicts user-item interactions based on both items’ interaction history and item content information. |
| Approach: | They propose to combine item encodings with a multi-modality approach to improve training efficiency by 146x . |
| Outcome: | The proposed model improves training efficiency (up to 146x) on five datasets from two task domains of Knowledge Tracing and News Recommendation. |
Discrete Prompt Optimization via Constrained Generation for Zero-shot Re-ranker (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing studies suggest re-rankers by fine-tuning pre-trained language models . however, manual search for discrete prompts is expensive and sub-optimal in transferability . |
| Approach: | They propose a discrete prompt optimization method that guides the generated texts toward optimal prompts . they propose to use large-scale language models as a zero-shot re-ranker . |
| Outcome: | The proposed method improves the performance of the re-ranker against baselines and human prompts. |
Joint Constrained Learning with Boundary-adjusting for Emotion-Cause Pair Extraction (2023.acl-long)
Copied to clipboard
| Challenge: | Emotion-Cause Pair Extraction (ECPE) aims to identify the document’s emotion clauses and corresponding cause clauses. |
| Approach: | They propose a constrained learning framework with boundary-adjusting for Emotion-Cause Pair Extraction that summarizes prior rules and forces the model to take them into consideration in optimization. |
| Outcome: | The proposed framework achieves competitive results compared with state-of-the-art methods on unbalanced data and proves robustness on unbalancing data. |
CondenseLM: LLMs-driven Text Dataset Condensation via Reward Matching (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for dataset condensation struggle to compress more information into samples . current methods struggle to extract enough training data for effective learning . |
| Approach: | They propose a paradigm for dataset condensation that uses an LLMs-driven approach to generate more informative and less biased samples. |
| Outcome: | The proposed method outperforms coreset selection and existing condensation methods by large margins while significantly reducing the computational cost. |
Harmonizing Dense and Sparse Signals in Multi-turn RL: Dual-Horizon Credit Assignment for Industrial Sales Agents (2026.acl-industry)
Copied to clipboard
| Challenge: | Large language models for industrial sales require balancing long-term commercial objectives with immediate linguistic constraints such as fluency and compliance. |
| Approach: | They propose a framework that disentangles optimization across time scales by normalizing advantages from turn-level and session-level rewards before fusion. |
| Outcome: | The proposed framework outperforms the state-of-the-art GRPO model in conversion rate and identity detection rate. |
Survival of the Safest: Towards Secure Prompt Optimization through Interleaved Multi-Objective Evolution (2024.emnlp-industry)
Copied to clipboard
| Challenge: | Experimental evaluations across diverse benchmark datasets affirm ‘s efficacy in delivering high performance and notably enhancing safety and security compared to single-objective methods. |
| Approach: | They propose a multi-objective prompt optimization framework that integrates semantic, feedback, and crossover mutations to effectively traverse the prompt landscape. |
| Outcome: | The proposed framework enhances both performance and security in large language models while keeping computational demands low. |
Fine-Tuning Large Language Models for Stock Return Prediction Using Newsflow (2024.emnlp-industry)
Copied to clipboard
| Challenge: | Quantitative investing relies on extracting quantitative features or signals from various data sources including market prices, economic indicators, financial text, etc. |
| Approach: | They propose to integrate LLMs’ token-level embeddings into a forecasting module and compare their results to those of encoder-only and decoder-based models. |
| Outcome: | The proposed model outperforms conventional sentiment scores on multiple investment universes and is based on encoder-only and decoder-based models. |
Exploiting Curriculum Learning in Unsupervised Neural Machine Translation (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Experimental results show that the proposed method achieves consistent improvements with faster convergence speed. |
| Approach: | They propose a curriculum learning method to gradually utilize pseudo bi-texts based on their quality from multiple granularities. |
| Outcome: | The proposed method achieves consistent improvements with faster convergence speed on WMT 14 En-Fr, WMT14 En-De, and LDC En-Zh translation tasks. |
Bridging Languages through Images with Deep Partial Canonical Correlation Analysis (P18-1)
Copied to clipboard
| Challenge: | a deep neural network can be used to improve bilingual text embeddings . a novel approach is proposed to optimize text embed-ings on shared visual information . |
| Approach: | They propose a deep neural network that leverages images to improve bilingual text embeddings. |
| Outcome: | The proposed model outperforms previous methods on word similarity and cross-lingual image description retrieval. |
Improving Visual-Semantic Embedding with Adaptive Pooling and Optimization Objective (2023.eacl-main)
Copied to clipboard
Zijian Zhang, Chang Shu, Ya Xiao, Yuan Shen, Di Zhu, Youxin Chen, Jing Xiao, Jey Han Lau, Qian Zhang, Zheng Lu
| Challenge: | Recent VSE models combine simple pooling methods with hard triplet loss to improve performance. |
| Approach: | They propose an adaptive pooling strategy that allows the model to learn how to aggregate features through a combination of simple pooling methods. |
| Outcome: | The proposed strategy outperforms current state-of-the-art systems on image-to-text and text-toimage retrieval. |
Beyond Discrete Search: Divergent Thinking as Intention Optimization in Latent Space (2026.acl-srw)
Copied to clipboard
| Challenge: | Despite rapid progress in LLM-based code generation, a persistent gap remains between what models can solve and what they solve on a given attempt. |
| Approach: | They propose a framework that recasts coding as optimization overconditioning contexts that influence the generation of natural-languagesolution intentions. |
| Outcome: | The proposed framework raises resolution rate of weak, quantized 24B open-weight model to parity with frontier models +25 its size. |
Knowledge Graph Embedding by Adaptive Limit Scoring Loss Using Dynamic Weighting Strategy (2022.findings-acl)
Copied to clipboard
| Challenge: | Existing knowledge graph embedding models use a loss framework to distinguish between correct and incorrect triplets. |
| Approach: | They propose a loss framework that reweights each triplet to highlight the less-optimized triplets. |
| Outcome: | The proposed method performs on several knowledge graph embedding models, including TransE, TransH and ComplEx. |
SwarmAgentic: Towards Fully Automated Agentic System Generation via Swarm Intelligence (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing agentic system generation frameworks lack autonomy, autonomy, and functionality . current frameworks are too rigid, limiting adaptability and scalability. |
| Approach: | They propose a framework that fully automates agentic system generation, optimization, and collaboration . they construct agents from scratch and jointly refine functionality and coordination . |
| Outcome: | The proposed framework outperforms ADAS on six real-world, open-ended, and exploratory tasks on the TravelPlanner benchmark. |
Revisiting OPRO: The Limitations of Small-Scale LLMs as Optimizers (2024.findings-acl)
Copied to clipboard
| Challenge: | Recent studies aim to enhance the efficacy of Large Language Models (LLMs) through strategic prompting. |
| Approach: | They propose to revisit the optimization by prompting approach for small-scale LLMs . they suggest future prompting engineering to consider both model capabilities and computational costs . |
| Outcome: | The proposed approach shows limited effectiveness in small-scale LLMs, with limited inference capabilities constraining optimization ability. |
Structured Pruning Learns Compact and Accurate Models (2022.acl-long)
Copied to clipboard
| Challenge: | Pre-trained language models have high costs in terms of storage, memory, and computation time. |
| Approach: | They propose a task-specific structured pruning method CoFi which provides highly parallelizable subnetworks and matches distillation methods in both accuracy and latency. |
| Outcome: | The proposed method matches the distillation methods in accuracy and latency without resorting to unlabeled data. |
A Wolf in Sheep’s Clothing: Generalized Nested Jailbreak Prompts can Fool Large Language Models Easily (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing methods for generating 'jailbreaks' suffer from manual design or require optimization on other white-box models, which compromises either generalization or efficiency. |
| Approach: | They propose a framework that leverages LLMs to generate effective jailbreak prompts and a generalized framework that can be used to generate prompts. |
| Outcome: | The proposed framework improves the attack success rate while reducing the time cost compared to baselines. |
Think Outside the Policy: In-Context Steered Policy Optimization (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing Reinforcement Learning from Verifiable Rewards (RLVR) methods exhibit limited exploration due to reliance on on-policy rollouts which are limited to the current policy’s distribution, resulting in narrow trajectory diversity. |
| Approach: | They propose a framework that leverages the in-context learning capability of Large Reasoning Models to provide expert guidance using existing datasets. |
| Outcome: | The proposed framework improves RLVR performance and training stability on mathematical reasoning benchmarks. |
Direct Prompt Optimization with Continuous Representations (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods for prompt optimization for language models lack extensibility and search space. |
| Approach: | They propose a method that integrates greedy strategies into optimization with continuous representations to address instability caused by rounding. |
| Outcome: | The proposed approach can improve prompt optimization performance on text classification and attack tasks, as well as models, including GPT-2, OPT, Vicuna, and LLaMA-2. |
Beyond Instruction Optimization: Multi-Agent Error-Driven Class Description Refinement for LLM-Based Classification (2026.acl-industry)
Copied to clipboard
| Challenge: | Large Language Models have demonstrated considerable efficacy in classification tasks . however, their performance depends on two critical prompt components: Task Instructions (HOW to classify) and Class Descriptions (WHAT defines each class). |
| Approach: | They propose a multi-agent framework for iteratively refining class descriptions based on classification errors. |
| Outcome: | Empirical evaluation shows up to 20.71% accuracy improvements over static class descriptions. |
Revisiting Self-Play Preference Optimization: On the Role of Prompt Difficulty (2026.findings-acl)
Copied to clipboard
| Challenge: | incorporating difficult prompts into training fails to enhance overall performance, e.g., as prompt difficulty decreases. |
| Approach: | They investigate how prompts of varying difficulty influence self-play preference optimization . they use the reward of sampled responses of a prompt as a proxy for its difficulty . |
| Outcome: | The proposed model improves on difficult prompts and easy prompts, but fails to train on difficult ones and learns from failures. |
Domain Adaptation for Question Answering via Question Classification (2022.coling-1)
Copied to clipboard
| Challenge: | Question answering systems often experience performance deterioration upon user-generated questions. |
| Approach: | They propose a question classification framework to help QA domains adapt to different domains. |
| Outcome: | The proposed framework improves on state-of-the-art datasets against multiple datasets. |
Learning Retrieval Augmentation for Personalized Dialogue Generation (2023.emnlp-main)
Copied to clipboard
| Challenge: | Personalized dialogue generation is a popular approach for conversational AI applications . however, persona profiles may not provide comprehensive descriptions of the persona . |
| Approach: | They propose a method that leverages persona profiles and dialogue context to generate personalized dialogues by leveraging personas and persona profile. |
| Outcome: | The proposed method outperforms baselines on the CONVAI2 dataset . it is expected to generate personalized dialogues based on persona profiles and dialogue context . |
GOODLIAR: A Reinforcement Learning-Based Deceptive Agent for Disrupting LLM Beliefs on Foundational Principles (2025.findings-acl)
Copied to clipboard
| Challenge: | Recent advances indicate that LLMs exhibit increasingly complex reasoning abilities . |
| Approach: | They propose a reinforcement learning framework that generates deceptive contexts to rewrite an LLM’s core axiomatic beliefs. |
| Outcome: | The proposed framework induces persistent belief shifts rather than one-off policy breaches. |
Analyzing LLM Instruction Optimization for Tabular Fact Verification (2026.findings-eacl)
Copied to clipboard
Xiaotang Du, Giwon Hong, Wai-Chung Kwan, Rohit Saxena, Ivan Titov, Pasquale Minervini, Emily Allaway
| Challenge: | evaluating instruction optimization for tabular fact verification is a key challenge for reliable NLP systems. |
| Approach: | They compare instruction optimization for tabular fact verification with a framework based on DSPy . they find that instruction optimization consistently improves verification accuracy . |
| Outcome: | The proposed method improves verification accuracy across four benchmarks and three model families. |
Optimizing Deeper Transformers on Small Datasets (2021.acl-long)
Copied to clipboard
Peng Xu, Dhruv Kumar, Wei Yang, Wenjie Zi, Keyi Tang, Chenyang Huang, Jackie Chi Kit Cheung, Simon J.D. Prince, Yanshuai Cao
| Challenge: | a common belief that training deep transformers from scratch requires large datasets is wrong . however, with proper initialization and optimization, the benefits of very deep transformer can carry over to challenging tasks with small datasets. |
| Approach: | They train 48 layers of transformers from pre-trained RoBERTa and 24 relation-aware layers from scratch. |
| Outcome: | The proposed scheme achieves state-of-the-art performance on a text-to-sql parsing benchmark . it uses 24 fine-tuned layers from pre-trained RoBERTa and 24 relation-aware layers from scratch . |
Imbalanced Gradients in RL Post-Training of Multi-Task LLMs (2026.findings-eacl)
Copied to clipboard
Runzhe Wu, Ankur Samanta, Ayush Jain, Scott Fujimoto, Jeongyeol Kwon, Ben Kretzu, Youliang Yu, Kaveh Hassani, Boris Vidolov, Yonathan Efroni
| Challenge: | Large-gradient tasks can achieve similar or even much lower learning gains than small-grading ones. |
| Approach: | They show that large-gradient tasks can achieve lower learning gains than small-grading ones . large-grade tasks can accomplish similar or even lower learning gain than small grade ones if they are large . |
| Outcome: | The proposed approach fails when certain tasks produce larger gradients . Large-gradient tasks can achieve lower learning gains than small-gradent ones . |
Compressing then Matching: An Efficient Pre-training Paradigm for Multimodal Embedding (2026.acl-long)
Copied to clipboard
Da Li, Yuxiao Luo, Keping Bi, Jiafeng Guo, Wei Yuan, Biao Yang, Yan Wang, Fan Yang, Tingting Gao, Guorui Zhou
| Challenge: | Recent approaches demonstrate that MLLMs can be adapted into competitive embedding models via large-scale contrastive learning. |
| Approach: | They propose a compressed pre-training phase which serves as a warm-up stage for contrastive learning. |
| Outcome: | The proposed model achieves state-of-the-art among MLLMs of comparable size on the MMEB, realizing optimization in both efficiency and effectiveness. |
Better Pre-Training by Reducing Representation Confusion (2023.findings-eacl)
Copied to clipboard
| Challenge: | Existing methods to improve pre-trained language models address information confusion in position encoding and model representations. |
| Approach: | They propose two techniques to improve pre-trained language models by decoupling directions and auxiliary regularizers. |
| Outcome: | The proposed techniques can improve pre-trained language models on GLUE benchmarks. |
Memory Dial: A Training Framework for Controllable Memorization in Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to memorization detection are post-hoc . large language models can reproduce training data verbatim, complicating accuracy estimates . |
| Approach: | They propose a training framework that makes memorization an explicit variable. |
| Outcome: | The proposed framework produces models identical in architecture, data, and optimization, but varying in memorization pressure. |
MAGMA – Multimodal Augmentation of Generative Models through Adapter-based Finetuning (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Large-scale pretraining is becoming the norm in Vision-Language (VL) modeling. |
| Approach: | They propose a method for augmenting generative language models with additional modalities using adapter-based finetuning. |
| Outcome: | The proposed method outperforms Frozen on open-ended generative tasks while maintaining the language model weights. |
SOS-LoRA: Static Orthogonal-Subspace Low-Rank Adaptation with Fixed Multi-Scale Scaling (2026.acl-long)
Copied to clipboard
| Challenge: | Low-Rank Adaptation (LoRA) is a parameter-efficient fine-tuning method for large language models. |
| Approach: | They propose a drop-in extension that reparameterizes a rank-rtot update as a sum of K *static* low-rank experts. |
| Outcome: | Experiments on reasoning and knowledge-intensive benchmarks show consistent gains over matched-budget LoRA. |
Towards Fast and Accurate Neural Chinese Word Segmentation with Multi-Criteria Learning (2020.coling-main)
Copied to clipboard
| Challenge: | Chinese word segmentation datasets have ambiguous annotation criteria resulting in multi-grained compounds. |
| Approach: | They propose a domain adaptive segmenter to exploit diverse annotation criteria of datasets . they use bidirectional encoder representations from transformers to introduce open-domain knowledge . |
| Outcome: | The proposed model outperforms the state-of-the-art models on 10 Chinese word datasets with superior efficiency. |
Stand on The Shoulders of Giants: Building JailExpert from Previous Attack Experience (2025.emnlp-main)
Copied to clipboard
Xi Wang, Songlei Jian, Shasha Li, Xiaopeng Li, Bin Ji, Ma Jun, Xiaodong Liu, Jing Wang, Jianfeng Zhang, Jie Yu, Feilong Bao, null Wangbaosheng
| Challenge: | Existing methods to generate human-aligned content with a “jailbreak prompt” are inefficient and repetitive, causing inefficiency and a lack of experience. |
| Approach: | They propose a framework that integrates past attack experiences to aid current jailbreak attempts. |
| Outcome: | The proposed framework improves both attack effectiveness and efficiency compared to the current black-box jailbreak method. |
SmartBench: Is Your LLM Truly a Good Chinese Smartphone Assistant? (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing evaluation benchmarks for Large Language Models focus on objective tasks like mathematics and coding in English, which do not reflect the practical use cases of on-device LLMs in real-world mobile scenarios. |
| Approach: | They propose a benchmark to evaluate the capabilities of on-device Large Language Models in Chinese mobile contexts. |
| Outcome: | The proposed framework evaluates on-device LLMs and MLLMs in Chinese . it provides a standardized framework for evaluating LLM performance on real smartphones . |
Zer0-Jack: A memory-efficient gradient-based jailbreaking method for black box Multi-modal Large Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Multi-modal large language models are highly vulnerable to jailbreak attacks due to their additional modality. |
| Approach: | They propose a black-box jailbreak framework based on zeroth-order optimization . they propose generating malicious images and patch-wise block coordinate descent . |
| Outcome: | The proposed framework achieves 98.2% success on MiniGPT-4 and 95% on the Harmful Behaviors Multi-modal dataset while jailbreaking commercial models such as GPT-4o. |
A Novel Two-step Fine-tuning Framework for Transfer Learning in Low-Resource Neural Machine Translation (2024.findings-naacl)
Copied to clipboard
| Challenge: | Existing transfer learning methods for neural machine translation use a well-trained translation model to initialize a child model with corresponding datasets. |
| Approach: | They propose a two-step fine-tuning framework for transfer learning in low-resource neural machine translation that adjusts the parent model to fit the child language by using the child source data. |
| Outcome: | The proposed framework improves on five low-resource translations on high-resolution languages. |
A Probabilistic Framework for Discovering New Intents (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods for detecting unknown intents do not explore the intrinsic structure of unlabeled data. |
| Approach: | They propose a probabilistic framework for discovering intents where intent assignments are treated as latent variables. |
| Outcome: | The proposed framework can be used to discover intents with latent variables . it can be applied to three challenging real-world datasets . |
Silencing the Guardrails: Inference-Time Jailbreaking via Dynamic Contextual Representation Ablation (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing strategies to circumvent safety constraints face significant trade-offs between effectiveness and efficiency. |
| Approach: | They propose a framework that allows to infer model refusal behaviors without expensive parameter updates or training. |
| Outcome: | The proposed framework outperforms baselines in multiple safety-aligned open-source LLMs. |
Should You Mask 15% in Masked Language Modeling? (2023.eacl-main)
Copied to clipboard
| Challenge: | Masked language models (MLMs) traditionally mask 15% of tokens due to the belief that more masking would leave insufficient context to learn good representations. |
| Approach: | They revisit the 15% masking rate of MLMs to examine the role of masking in linguistic training. |
| Outcome: | The proposed masking rate outperforms BERT-large size models on GLUE and SQUAD while maintaining 95% accuracy. |
Revitalizing Black-Box Interpretability: Actionable Interpretability for LLMs via Proxy Models (2026.acl-long)
Copied to clipboard
| Challenge: | Applying model-agnostic explanations to Large Language Models is hindered by prohibitive computational costs rendering them dormant for real-world applications. |
| Approach: | They propose a budget-friendly proxy framework that leverages efficient models to approximate the decision boundaries of expensive Large Language Models. |
| Outcome: | The proposed framework achieves over 90% fidelity with only 9.5% of the oracle’s cost and is open-source to facilitate future research. |
Sharpness-Aware Minimization for Topic Models with High-Quality Document Representations (2025.naacl-long)
Copied to clipboard
Tung Nguyen, Tue Le, Hoang Tran Vuong, Quang Duc Nguyen, Duc Anh Nguyen, Linh Ngo Van, Sang Dinh, Thien Huu Nguyen
| Challenge: | Recent advances in topic models have significantly enhanced the performance compared to conventional probabilistic approaches. |
| Approach: | They propose to use a topic model optimization method to improve generalization and performance. |
| Outcome: | The proposed method improves topic models on popular benchmark datasets. |
LeeBERT: Learned Early Exit for BERT with cross-level optimization (2021.acl-long)
Copied to clipboard
| Challenge: | Pre-trained language models are resource exhaustive and computationally expensive for industrial scenarios. |
| Approach: | They propose a learning scheme to learn from each other to speed up inference . they ask each exit to learn the weights of different loss terms, instead of learning only from the last layer . |
| Outcome: | The proposed scheme improves state-of-the-art (SOTA) early exit methods for pre-trained models on the GLUE benchmark. |
Can Multimodal LLMs See Materials Clearly? A Multimodal Benchmark on Materials Characterization (2025.findings-emnlp)
Copied to clipboard
Zhengzhao Lai, Youbin Zheng, Zhenyang Cai, Haonan Lyu, Jingpu Yang, Hong-Qing Liang, Yan Hu, Benyou Wang
| Challenge: | characterization imaging data is fundamental to acquiring materials information. |
| Approach: | a team of researchers develop a benchmark for materials characterization imaging data . the goal is to bridge this gap by addressing 1,500 questions that require expert-level expertise. |
| Outcome: | a new benchmark for materials characterization imaging data is presented . the benchmark reveals that MLLMs perform poorly when addressing higher-level questions . |
Different Tunes Played with Equal Skill: Exploring a Unified Optimization Subspace for Parameter-Efficient Tuning (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Existing delta tuning algorithms freeze most of the parameters and only optimize minimal adaptive parameters. |
| Approach: | They propose to decompose DETs into a unified optimization subspace and conduct optimization within the subspace. |
| Outcome: | The proposed DETs achieve comparable performance to the original DET and can be transferred to another DET with non-trivial performance. |
Incentivizing Parametric Knowledge via Reinforcement Learning with Verifiable Rewards for Cross-Cultural Entity Translation (2026.acl-long)
Copied to clipboard
Jiang Zhou, Xiaohu Zhao, Xinwei Wu, Tianyu Dong, Hao Wang, Yangyang Liu, Heng Liu, Linlong Xu, Longyue Wang, Weihua Luo, Deyi Xiong
| Challenge: | Current systems often fall short of this goal in settings where translation hinges on culturally grounded entities such as books, films, places, songs and idioms. |
| Approach: | They propose a framework that anchors supervision on a verifiable, entity-level reward signal and incorporates lightweight structural gates to stabilize optimization. |
| Outcome: | The proposed framework improves on XC-Translate and shows that it can learn a robust reasoning process rather than imitating reference translations. |
GradOT: Training-free Gradient-preserving Offsite-tuning for Large Language Models (2025.acl-long)
Copied to clipboard
Kai Yao, Zhaorui Tan, Penglei Gao, Lichun Li, Kaixin Wu, Yinggui Wang, Yuan Zhao, Yixin Ji, Jianke Zhu, Wei Wang
| Challenge: | Existing methods for offsite-tuning of large language models require high computational costs and lack theoretical analysis. |
| Approach: | They propose an offsite-tuning approach that selectively applies compression techniques such as rank compression and channel pruning to preserve the gradients of fine-tuned adapters while ensuring privacy. |
| Outcome: | The proposed method surpasses existing OT methods in privacy protection and model performance. |
Continually Detection, Rapidly React: Unseen Rumors Detection Based on Continual Prompt-Tuning (2022.coling-1)
Copied to clipboard
| Challenge: | Existing rumor detection models assume the same training and testing distributions and can not cope with the continuously changing social network environment. |
| Approach: | They propose a Continual Prompt-Tuning RD framework which avoids catastrophic forgetting of upstream tasks during sequential task learning and enables bidirectional knowledge transfer between domain tasks. |
| Outcome: | The proposed framework avoids catastrophic forgetting (CF) of upstream tasks during sequential task learning and enables bidirectional knowledge transfer between domain tasks. |
Understanding Conflicts in Multi-Objective Alignment through Reward Consistency (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing training pipelines still face alignment conflicts where optimizing for one objective degrades performance on others. |
| Approach: | They propose a reward-based criterion that approximates alignment conflicts via reward models. |
| Outcome: | The proposed framework improves harmlessness and helpfulness scores by 23.07% over the vanilla dataset. |
Heterogeneous Adaptive Policy Optimization: Tailoring Optimization to Every Token’s Nature (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods that use entropy as a discrete filter or post-hoc regulator are limited in their ability to optimize for reasoning tasks. |
| Approach: | They propose a token-aware algorithm that continuously adapts optimization dynamics based on token-level entropy throughout the entire training process. |
| Outcome: | Extensive experiments on mathematical reasoning, code, and logic tasks across multiple models demonstrate HAPO’s consistent superiority over DAPO. |
Can LLMs Clarify? Investigation and Enhancement of Large Language Models on Argument Claim Optimization (2025.coling-main)
Copied to clipboard
| Challenge: | While Large Language Models (LLMs) have demonstrated proficiency in text rewriting tasks such as style transfer and query rewrite, their application to claim optimization remains unexplored. |
| Approach: | They propose to use a sliding window mechanism to evaluate the performance of large language models in claim clarification tasks under different settings. |
| Outcome: | The proposed model improves the performance of three LLMs on the claim clarification task under zero-shot, few-shot and supervised fine-tuning settings. |
Efficient Transformer-based Large Scale Language Representations using Hardware-friendly Block Structured Pruning (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Pretrained large-scale language models have been criticized for their limited weight storage and computational speed on hardware platforms. |
| Approach: | They propose an efficient transformer-based large-scale language representation using hardware-friendly block structure pruning. |
| Outcome: | The proposed model achieves 5.0x accuracy on GLUE benchmarks and 1.79x compression rate on DistilBERT. |
What Makes a Good Natural Language Prompt? (2025.acl-long)
Copied to clipboard
| Challenge: | Existing studies on prompt quality show imbalanced support across models and tasks, and research gaps. |
| Approach: | They propose a property- and human-centric framework for evaluating prompt quality . they propose comparing prompt quality to other factors such as adverbs and apverbs . |
| Outcome: | The proposed framework reveals imbalanced support across models and tasks and substantial research gaps. |
LATENTLOGIC: Learning Logic Rules in Latent Space over Knowledge Graphs (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for learning logic rules for knowledge graph reasoning face limitations such as searching in vast search space and inefficient optimization. |
| Approach: | They propose a framework to efficiently mine logic rules by controllable generation in the latent space by a pre-trained VAE and a discriminator. |
| Outcome: | The proposed framework efficiently mines logic rules by controllable generation in the latent space. |
Modal Feature Optimization Network with Prompt for Multimodal Sentiment Analysis (2025.coling-main)
Copied to clipboard
| Challenge: | Multimodal sentiment analysis(MSA) is used to understand human emotional states through multimodal. |
| Approach: | They propose a Modal Feature Optimization Network with a modal prompt attention mechanism to optimize the under-optimized modal representation by determining which modalities are under- optimized . |
| Outcome: | The proposed method outperforms existing state-of-the-art models on public benchmark datasets. |
On the Encoder-Decoder Incompatibility in Variational Text Modeling and Beyond (2020.acl-main)
Copied to clipboard
| Challenge: | Existing work has shown that the optimization of variational autoencoders suffers from the posterior collapse problem. |
| Approach: | They propose a variational autoencoder that couples a VAE model with a deterministic autoencoding model and improves the parameters via weight sharing and decoder signal matching. |
| Outcome: | The proposed model improves on benchmark datasets and improves diversity of dialogue generation. |
Rethinking Skip Connection with Layer Normalization (2020.coling-main)
Copied to clipboard
| Challenge: | Existing methods to solve the optimization problem of deep neural networks are not linear, but can be used as a modulating mechanism between the input and output. |
| Approach: | They propose to use skip connection to adjust the scale of the input and output to improve the performance. |
| Outcome: | The proposed approach improves performance and convergence of deep neural networks and can be applied to machine translation and image classification datasets. |
Meta-Transfer Learning for Code-Switched Speech Recognition (2020.acl-main)
Copied to clipboard
| Challenge: | Increasing number of people in the world today speak a mixed-language as a result of being multilingual. |
| Approach: | They propose a method to transfer learn on a code-switched speech recognition system by extracting information from high-resource monolingual datasets. |
| Outcome: | The proposed model outperforms baselines on speech recognition and language modeling tasks and is faster to converge. |
To Copy Rather Than Memorize: A Vertical Learning Paradigm for Knowledge Graph Completion (2023.acl-long)
Copied to clipboard
Rui Li, Xu Chen, Chaozhuo Li, Yanming Shen, Jianan Zhao, Yujing Wang, Weihao Han, Hao Sun, Weiwei Deng, Qi Zhang, Xing Xie
| Challenge: | Existing methods for embedding knowledge graphs implicitly memorize relation rules to infer missing links, but they are difficult to memorize due to the inherent deficiencies of such implicit memorization strategy. |
| Approach: | They propose a vertical learning paradigm that allows to explicitly copy target information from related factual triples for more accurate prediction. |
| Outcome: | The proposed model improves generalization ability and makes distant link prediction significantly easier. |
LinkPrompt: Natural and Universal Adversarial Attacks on Prompt-based Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Prompt-based learning is a new language model training paradigm that adapts Pre-trained Language Models (PLMs) to downstream tasks. |
| Approach: | They propose a prompt-based learning paradigm that adapts Pre-trained Language Models to downstream tasks . they use a gradient-based beam search algorithm to generate adversarial triggers . |
| Outcome: | The proposed model improves performance on various natural language processing tasks by optimizing the prompt template. |
Engagement Undermines Safety: How Stereotypes and Toxicity Shape Humor in Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Large language models are increasingly used for creative writing and engagement content, raising safety concerns about their outputs. |
| Approach: | They evaluate how funniness optimization in large language models couples with harmful content by jointly measuring humor, stereotypicality, and toxicity. |
| Outcome: | The proposed model couples humor, stereotypicality, and toxicity with harmful outputs . the results suggest a bias amplification loop between generators and evaluators . |
A Survey of Large Language Model-Based Search Agents (2026.acl-long)
Copied to clipboard
Yunjia Xi, Jianghao Lin, Yongzhao Xiao, Zheli Zhou, Rong Shan, Te Gao, Jiachen Zhu, Weiwen Liu, Yong Yu, Weinan Zhang
| Challenge: | Large Language Models (LLMs) have revolutionized web search, but their integration is static and cannot handle complex contexts. |
| Approach: | They analyze existing research and analyze existing work from the perspectives of architecture, optimization, application, and evaluation. |
| Outcome: | The proposed models can comprehend user intentions and context and execute multi-turn retrieval with dynamic planning, extending search capabilities far beyond the web. |
AGrail: A Lifelong Agent Guardrail with Effective and Adaptive Safety Detection (2025.acl-long)
Copied to clipboard
| Challenge: | Existing defense agencies fail to adaptively and effectively mitigate these risks. |
| Approach: | They propose a lifelong agent guardrail that enhances LLM agent safety by enabling adaptive safety check generation, effective safety check optimization, and tool compatibility & flexibility. |
| Outcome: | The proposed agent guardrail achieves strong performance against task-specific and systemic risks and is transferable across different LLM agents’ tasks. |
Learning to Instruct: Fine-Tuning a Task-Aware Instruction Optimizer for Black-Box LLMs (2025.findings-emnlp)
Copied to clipboard
Yunzhe Qi, Jinjin Tian, Tianci Liu, Ruirui Li, Tianxin Wei, Hui Liu, Xianfeng Tang, Monica Xiao Cheng, Jingrui He
| Challenge: | Learning to Instruct is a new paradigm for black-box LLMs with inaccessible internal states. |
| Approach: | They propose a new paradigm that formulates instruction optimization as an LLM fine-tuning objective for a white-box “instruction engineer” LLM. |
| Outcome: | The proposed framework outperforms strong baselines in performance and efficiency. |
Visualizing and Understanding the Effectiveness of BERT (D19-1)
Copied to clipboard
| Challenge: | Language model pre-training, such as BERT, has achieved strong performance in many NLP tasks. |
| Approach: | They propose to visualize loss landscapes and optimization trajectories of fine-tuning BERT on specific datasets. |
| Outcome: | The proposed model improves performance and generalization capability across tasks. |
IHEval: Evaluating Language Models on Following the Instruction Hierarchy (2025.naacl-long)
Copied to clipboard
Zhihan Zhang, Shiyang Li, Zixuan Zhang, Xin Liu, Haoming Jiang, Xianfeng Tang, Yifan Gao, Zheng Li, Haodong Wang, Zhaoxuan Tan, Yichuan Li, Qingyu Yin, Bing Yin, Meng Jiang
| Challenge: | Instruction-tuned language models (LMs) are increasingly deployed as interactive services across various applications. |
| Approach: | They propose a benchmark to evaluate models' ability to follow the instruction hierarchy by comparing their models to a set of benchmarks. |
| Outcome: | The proposed benchmark covers 3,538 examples across nine tasks covering cases where instructions in different priorities either align or conflict. |
Beyond BLEU:Training Neural Machine Translation with Semantic Similarity (P19-1)
Copied to clipboard
| Challenge: | Recent work has shown that optimizing neural machine translation systems to directly improve evaluation metrics such as BLEU can improve final translation accuracy. |
| Approach: | They propose a reward function that assigns partial credit to BLEU and provides more diversity in scores than BLUE. |
| Outcome: | The proposed reward function improves translation accuracy, semantic similarity, and human evaluation on four languages trans-lated to English and the optimization procedure converges faster. |
Retrieval-Augmented Few-shot Text Classification (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for retrieval-augmented text classification are successful in the few-shot scenario with limited retrieval space. |
| Approach: | They propose to use EM-L and R-L to provide task-specific guidance to retrieval metric . they also propose to incorporate retrieved memory alongside parameters for better generalization . |
| Outcome: | The proposed methods perform better on the few-shot scenario with limited retrieval space. |
Direct Metric Optimization for Image Captioning through Reward-Weighted Augmented Data Utilization (2024.acl-long)
Copied to clipboard
Takumi Takada, Yuma Suzuki, Hiroki Takushima, Hayato Tanoue, Haruki Sato, Aiswariya Kumar, Hiroki Nishihara, Takayuki Hori, Kazuya Ueki
| Challenge: | Recent large-scale vision language models (VLMs) lack continuity between learning objective and performance metrics. |
| Approach: | They propose a lightweight final-metric-optimizing training method that replaces the expensive exploration process in RL with an offline, diverse text data augmentation method. |
| Outcome: | The proposed method achieves comparable performance to state-of-the-art RL method while saving hundreds of times more model forwarding iterations and greater amounts of computation time. |
Tucker Decomposition with Frequency Attention for Temporal Knowledge Graph Completion (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing models for temporal knowledge graph completion only consider the combination of one relation with one timestamp, ignoring the global nature of the embedding. |
| Approach: | They propose a temporal knowledge Graph Completion model that captures global temporal dependencies between one relation and the entire timestamp. |
| Outcome: | The proposed model outperforms the state-of-the-art models on three standard TKGC datasets on several metrics. |
Distributionally Robust Multilingual Machine Translation (2021.emnlp-main)
Copied to clipboard
| Challenge: | Multilingual neural machine translation (MNMT) learns to translate multiple language pairs with a single model, but the data imbalance hinders it from performing uniformly across language pairs. |
| Approach: | They propose a distributionally robust optimization objective which minimizes the worst-case expected loss over the set of language pairs. |
| Outcome: | The proposed learning objective outperforms baseline methods on three sets of languages and shows that it is cost-effective and efficient. |
VPTQ: Extreme Low-bit Vector Post-Training Quantization for Large Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent research has focused on pushing weight-only quantization to extremely low-bit due to numerical representation limitations. |
| Approach: | They propose a vector-based quantization approach that pushes LLMs to extremely low-bit . they propose scalar-based weight quantization that reduces memory requirements and optimizes storage costs . |
| Outcome: | The proposed method reduces model quantization perplexity by 0.01-0.34 on LLaMA-2, 0.38-0.68 on mistral-7B, 4.41-7.34, on llaMA-3 on QA tasks on average. |
Rewarding What Matters: Step-by-Step Reinforcement Learning for Task-Oriented Dialogue (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing RL methods focus on generation tasks while neglecting dialogue state tracking (DST) for understanding. |
| Approach: | They propose a method that integrates RL into both understanding and generation tasks by introducing step-by-step rewards throughout the token generation. |
| Outcome: | The proposed approach achieves state-of-the-art results on three widely used datasets. |
Morfessor EM+Prune: Improved Subword Segmentation with Expectation Maximization and Pruning (2020.lrec-1)
Copied to clipboard
| Challenge: | Subword segmentation is a standard preprocessing step in many neural approaches to natural language processing. |
| Approach: | They propose to train a unigram subword model using a recursive algorithm and lexicon pruning algorithm. |
| Outcome: | The proposed method improves on the original training algorithm and improves morphological segmentation accuracy. |
AlgoPuzzleVQA: Diagnosing Multimodal Reasoning Challenges of Language Models with Algorithmic Multimodal Puzzles (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing datasets focused on visual question-answering focus on visual, language, and algorithmic knowledge . a new study examines the performance of multimodal language models in solving algorithmic puzzles . |
| Approach: | They propose a dataset to test the capabilities of multimodal language models in solving algorithmic puzzles. |
| Outcome: | The proposed dataset is generated automatically from human code. |
An Adaptive Logical Rule Embedding Model for Inductive Reasoning over Temporal Knowledge Graphs (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for temporal knowledge graphs (TKGs) are incomplete and therefore lack interpretability. |
| Approach: | They propose an interpretable temporal knowledge graph reasoning model that captures deep causal logic by learning rule embeddings. |
| Outcome: | The proposed model outperforms state-of-the-art models on the ICEWS14, ICEW0515 and ICEw18 datasets. |
EfficientQAT: Efficient Quantization-Aware Training for Large Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Quantization-aware training (QAT) is a low-bit training solution that requires substantial training resources. |
| Approach: | They propose an algorithm that reduces memory consumption by low-bit representations with minimal accuracy loss. |
| Outcome: | EfficientQAT achieves 2-bit Llama-2-70B model on single GPU in 41 hours . compared to previous methods, it obtains model with less than 3 points accuracy degradation . |
Sharpness-Aware Minimization Improves Language Model Generalization (2022.acl-long)
Copied to clipboard
| Challenge: | Comparatively little work has been done to improve the generalization of language models . recent work shows that Sharpness-Aware Minimization (SAM) can improve generalization without much computational overhead. |
| Approach: | They propose a Sharpness-Aware Minimization procedure that encourages convergence to flatter minima to improve generalization of language models without much computational overhead. |
| Outcome: | The proposed Sharpness-Aware Minimization procedure can improve language models without much computational overhead. |
SCOPE: Optimizing Key-Value Cache Compression in Long-context Generation (2025.acl-long)
Copied to clipboard
| Challenge: | Excessive compression during the prefill phase impairs comprehension of reasoning tasks . SCOPE is a framework that performs KV cache optimization during the decoding and prefill phases . |
| Approach: | They propose a framework that performs optimization during the prefill and decoding phases . they propose enabling a sliding strategy to select essential heavy hitters for the decoding phase . |
| Outcome: | Experiments show that SCOPE can optimize key-value cache for long-context generation tasks . the framework can preserve essential information while minimizing memory usage and transfer . |
MOA: Multi-Objective Alignment for Role-Playing Agents (2026.acl-long)
Copied to clipboard
| Challenge: | Prior work on role-playing agents relies on supervised fine-tuning or reinforcement learning with scalarized rewards, but these approaches do not address the coordination of multiple reward dimensions during optimization. |
| Approach: | They propose a reinforcement-learning framework that enables multi-dimensional, fine-grained rubric optimization for general RPAs. |
| Outcome: | Experiments on PersonaGym and RoleMRC show that MOA improves multi-dimensional role-playing performance over supervised and standard RL baselines. |
Document-Level Event Argument Extraction With a Chain Reasoning Paradigm (2023.acl-long)
Copied to clipboard
| Challenge: | Document-level event argument extraction aims to identify event arguments beyond sentence level, where a significant challenge is to model long-range dependencies. |
| Approach: | They propose a chain reasoning paradigm which captures long-range interdependence due to the chains’ compositional nature and generates decomposable first-order logic rules for reasoning. |
| Outcome: | The proposed method outperforms previous methods on two benchmarks and is robust enough to defend against adversarial attacks. |
Adaptive Prompt Structure Factorization: A Framework for Self-Discovering and Optimizing Compositional Prompt Programs (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) exhibit strong capabilities in reasoning, coding, and complex generation, yet their performance is highly sensitive to prompt design. |
| Approach: | They propose an API-only framework that decomposes a single prompt into semantic factors and updates selected factors while freezing the rest. |
| Outcome: | The proposed framework outperforms strong baselines, improves accuracy by up to +4.29 percentage points on average, and reduces optimization cost by 45–87% tokens on MultiArith while reaching peak validation in 1 step. |
MAPO: Advancing Multilingual Reasoning through Multilingual-Alignment-as-Preference Optimization (2024.acl-long)
Copied to clipboard
| Challenge: | Existing models exhibit inconsistent reasoning abilities across different languages . existing models lack consistency across languages due to imbalance of training data . |
| Approach: | They propose a multilingual alignment-as-preference optimization framework to align reasoning processes in other languages with the dominant language. |
| Outcome: | The proposed framework improves multilingual reasoning across languages on three benchmarks. |
Modeling Layout Reading Order as Ordering Relations for Visually-rich Document Understanding (2024.emnlp-main)
Copied to clipboard
Chong Zhang, Yi Tu, Yixi Zhao, Chenshu Yuan, Huan Chen, Yue Zhang, Mingxu Chai, Ya Guo, Huijia Zhu, Qi Zhang, Tao Gui
| Challenge: | Existing models of layout reading order do not convey the complete reading order information in the layout. |
| Approach: | They propose to model layout reading order as ordering relations over layout elements . they propose a reading-order-relation-enhancing pipeline to improve model performance . |
| Outcome: | The proposed model outperforms existing models on a visual-rich document dataset and on eight cross-domain VrD-IE/QA tasks without targeted optimization. |
AdamS: Momentum Itself Can Be A Normalizer for LLM Pretraining and Post-training (2025.emnlp-main)
Copied to clipboard
| Challenge: | Empirically, AdamS demonstrates strong performance in various tasks . et al., 2023b): AdamS is efficient, efficient, and model-agnostic. |
| Approach: | They propose a model-agnostic alternative to Adam for large language model pretraining and post-training. |
| Outcome: | The proposed method matches memory footprint of SGD with momentum while delivering superior performance. |
Q-Mamba: Towards more efficient Mamba models via post-training quantization (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing studies show that Mamba architectures have room for further optimization in linear projections and state caches. |
| Approach: | They propose a decoupled scale quantization scheme to mitigate outliers in states and channels by applying separate quantization scales. |
| Outcome: | The proposed method reduces memory consumption by 50% across various quantization settings, model sizes, and generation and zero-shot tasks. |
PRISM: Probabilistic Reward Model with Inherent Structural Modeling (2026.acl-long)
Copied to clipboard
| Challenge: | Existing evaluators compress diverse human judgments into a single scalar, leading to brittle alignment and reward hacking. |
| Approach: | They propose a Gaussian-based reinterpretation of reward evaluation as a conditional distribution and a mixture of Gaussians to capture conflicting preference dimensions. |
| Outcome: | The proposed model outperforms scalar baselines in accuracy and generalization. |
GenPTQ: Green Post-Training Quantization for Large-Scale ASR Models with Mixed-Precision Bit Allocation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Large-scale models with high memory and computation demands face significant challenges for deployment. |
| Approach: | They propose a weight-only quantization method that optimizes for weights and QAT to restore accuracy. |
| Outcome: | The proposed method achieves 89.1% model size reduction with only 0.8% increase in WER and completes optimization in just 15 seconds. |
Self-Improvement Towards Pareto Optimality: Mitigating Preference Conflicts in Multi-Objective Alignment (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to optimize large language models with human preferences suffer from preference conflicts in the data. |
| Approach: | They propose to construct Pareto-optimal responses to resolve preference conflicts by using a self-improving DPO framework that enables LLMs to self-generate and select Paret-optimized responses. |
| Outcome: | The proposed framework achieves superior Pareto Front performance over baselines on two datasets. |
TranSHER: Translating Knowledge Graph Embedding with Hyper-Ellipsoidal Restriction (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing knowledge graph embedding methods restrict entities on hyper-ellipsoid surfaces, resulting in suboptimal knowledge graph completion. |
| Approach: | They propose a score function that leverages relation-specific translations between head and tail entities to relax constraints on hyper-ellipsoid surfaces. |
| Outcome: | The proposed method achieves state-of-the-art performance on link prediction and generalizes well to datasets in different domains and scales. |
STEAM: A Semantic-Level Knowledge Editing Framework for Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for locate-and-editing focus on token-level likelihood optimization without addressing semantic coherence. |
| Approach: | They propose a semantic-level knowledge editing framework that enhances integration of updated knowledge into the model's knowledge structure. |
| Outcome: | The proposed framework improves integration of updated knowledge into the model's knowledge structure and improves semantic coherence. |
Leveraging Synthetic Targets for Machine Translation (2023.findings-acl)
Copied to clipboard
| Challenge: | Using synthetic target data, training models on synthetic targets outperforms training on actual ground-truth data. |
| Approach: | They propose a recipe for training machine translation models on synthetic target data by leveraging a large pre-trained model. |
| Outcome: | The proposed model outperforms training on real-world translation datasets. |
Benchmarking Chinese Commonsense Reasoning of LLMs: From Chinese-Specifics to Reasoning-Memorization Correlations (2024.acl-long)
Copied to clipboard
| Challenge: | Currently, many benchmarks evaluate the commonsense reasoning of large language models (LLMs), but most are English-based, limiting non-English evaluations. |
| Approach: | They propose to use Chinese commonsense reasoning to evaluate LLMs' commonsensing ability. |
| Outcome: | The proposed benchmark covers both globally known and Chinese-specific commonsense reasoning abilities and can be used as a reference for future research. |
Making Harmful Behaviors Unlearnable for Large Language Models (2024.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are often customized by fine-tuning for the requirements of different domains. |
| Approach: | They propose a controllable training framework to make undesired behaviors unlearnable during the fine-tuning process. |
| Outcome: | The proposed framework makes undesired behaviors unlearnable during the fine-tuning process while preserving the ability to learn other information. |
Reinforced Informativeness Optimization for Long-Form Retrieval-Augmented Generation (2026.acl-long)
Copied to clipboard
| Challenge: | Existing reinforcement learning systems lack verifiable reward mechanisms for long-form question answering . current systems lack reliable long-term answers due to lack of factual content . |
| Approach: | They propose a framework for reinforced verifiable informativeness optimization . it defines informativeness as measurable and externally verifier objective for RL . |
| Outcome: | Experiments show that RioRAG achieves higher factual recall and faithfulness . the proposed framework is based on a framework that uses nugget-centric verification with cross-source checks . |
ICG: Improving Cover Image Generation via MLLM-based Prompting and Personalized Preference Alignment (2025.emnlp-main)
Copied to clipboard
Zhipeng Bian, Jieming Zhu, Qijiong Liu, Wang Lin, Guohao Cai, Zhaocheng Du, Jiacheng Sun, Zhou Zhao, Zhenhua Dong
| Challenge: | Large language models and diffusion models have opened new possibilities for AI-generated content . personalized cover image generation remains underexplored despite its critical role in boosting user engagement on digital platforms. |
| Approach: | They propose a framework that integrates MLLM-based prompting with personalized preference alignment to generate high-quality, contextually relevant covers. |
| Outcome: | The proposed framework improves image quality, semantic fidelity, and personalization, leading to stronger user appeal and offline recommendation accuracy in downstream tasks. |
Making Every Step Effective: Jailbreaking Large Vision-Language Models Through Hierarchical KV Equalization (2025.findings-emnlp)
Copied to clipboard
| Challenge: | HKVE selectively accepts gradient optimization results based on the distribution of attention scores across different layers, ensuring that every optimization step positively contributes to the attack. |
| Approach: | They propose a framework that selectively accepts gradient optimization results based on the distribution of attention scores across different layers and selectively takes them into account when calculating the attack success rate. |
| Outcome: | The proposed framework outperforms existing methods by achieving success rates of 75.08% on MiniGPT4, 85.84% on LLaVA and 81.00% on Qwen-VL. |
RAISE: Reinforced Adaptive Instruction Selection For Large Language Models (2025.findings-emnlp)
Copied to clipboard
Qingsong Lv, Yangning Li, Zihua Lan, Zishan Xu, Jiwei Tang, Tingwei Lu, Yinghui Li, Wenhao Jiang, Hong-Gee Kim, Hai-Tao Zheng, Philip S. Yu
| Challenge: | Existing selection methods rely on static, heuristic quality scores and are executed only once before training. |
| Approach: | They propose a dynamic selection framework that integrates selection into every training step. |
| Outcome: | The proposed framework integrates selection into every training step. |
Few-shot Temporal Pruning Accelerates Diffusion Models for Text Generation (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing acceleration methods for text generation ignore the importance of the distribution of sampling steps, resulting in slow sampling rates. |
| Approach: | They propose a technique to accelerate diffusion models for text generation without additional training by using a Bayesian optimization approach. |
| Outcome: | The proposed technique achieves 400x acceleration even with minimal sampling steps after down to less than 1 minute of optimization yielding a competitive performance even with minimum sampling steps. |
Time Course MechInterp: Analyzing the Evolution of Components and Knowledge in Large Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Large language models acquire and store factual knowledge for interpretability, reliability, efficiency . prior work on factual recall focused on localizing knowledge within transformer parameters . |
| Approach: | They analyze the evolution of factual knowledge representation in a large language model by tracking its attention heads and feed forward networks over training. |
| Outcome: | The proposed model acquires and stores factual knowledge over time and is adaptively trained . the proposed model can be pruned, optimized, and transparent . |
Disentangle to Decay: Linear Attention with Trainable Decay Factor (2025.coling-main)
Copied to clipboard
| Challenge: | Existing linear attention models use a decay factor based positional encoding (PE), but the decay factor is manually designed and non-trainable, limiting further optimization. |
| Approach: | They propose a PE-based positional encoding that disentangles decay factor into two parts to achieve further optimization and stable training. |
| Outcome: | The proposed model achieves stable training of decay factor and improves inference efficiency in normal context and extrapolation scenarios. |
MDCR: A Dataset for Multi-Document Conditional Reasoning (2024.findings-emnlp)
Copied to clipboard
| Challenge: | ConditionalQA is limited to questions on single documents, neglecting harder cases that may require *cross-document reasoning* and *optimization*. |
| Approach: | They propose to use a dataset to evaluate models' ability to answer eligibility questions on single documents. |
| Outcome: | The proposed dataset can reflect real-world challenges and serve as a test bed for complex conditional reasoning that requires optimization. |
Differentiable Instruction Optimization for Cross-Task Generalization (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing studies have shown that instruction tuning is effective for generalizing to arbitrary tasks unseen during training. |
| Approach: | They propose to introduce learnable instructions and optimize them with gradient descent to optimize instruction for generalization ability. |
| Outcome: | The proposed instruction extractor extracts appropriate instruction and improves generalization ability compared to manual instruction tuning. |
Semantic Similarity Covariance Matrix Shrinkage (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods to estimate covariance matrix relied on historical price data and ignored company fundamental data. |
| Approach: | They propose to use semantic similarity to improve covariance estimations by using a shrinkage target. |
| Outcome: | The proposed method is compared with the prior art estimate for covariance shrinkage using semantic similarity and price history. |
PerfCoder: Large Language Models for Interpretable Code Performance Optimization (2026.findings-acl)
Copied to clipboard
Jiuding Yang, Shengyao Lu, Hongxuan Liu, Shayan Shirahmad Gale Bagi, Zahra Fazel, Tomasz Czajkowski, Di Niu
| Challenge: | Large language models (LLMs) have advanced automatic code generation, but their ability to produce high-performance code remains limited. |
| Approach: | They propose a family of large language models that generate performance-enhanced code through interpretable and customized optimization strategies. |
| Outcome: | The proposed model outperforms existing models on the PIE code performance benchmark and produces interpretable feedback that can guide larger LLMs in a planner–optimizer workflow. |
Adaptive Contrastive Learning on Multimodal Transformer for Review Helpfulness Prediction (2022.emnlp-main)
Copied to clipboard
| Challenge: | Modern review helpfulness prediction systems focus on polishing cross-modal representations and suffer from inferior optimization. |
| Approach: | They propose a method to polish cross-modal relation representations by learning mutual information through contrastive learning. |
| Outcome: | The proposed framework outperforms baselines and achieves state-of-the-art results on two publicly available datasets. |
FastMem: Fast Memorization of Prompt Improves Context Awareness of Large Language Models (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models struggle with context awareness, leading to inaccuracies in tasks requiring faithful adherence to provided information. |
| Approach: | They propose a method to enhance LLMs' context awareness by updating only the last Feed-Forward Network module to maximize the likelihood of the prompt before inference . |
| Outcome: | The proposed method improves the accuracy of Llama 3-8B-Inst on the NQ-SWAP dataset from 59.1% to 71.6% and reduces the output structure failure rate of Qwen 1.5-4B-Chat from 34.9% to 25.5%. |
Language-Coupled Reinforcement Learning for Multilingual Retrieval-Augmented Generation (2026.findings-acl)
Copied to clipboard
Rui Qi, Fengran Mo, Yufeng Chen, Xue Zhang, Shuo Wang, Hongliang Li, Xu Jinan, Meng Jiang, Jian-Yun Nie, Kaiyu Huang
| Challenge: | Existing approaches to multilingual retrieval-augmented generation (MRAG) use a single-turn retrieval and subsequent optimization to acquire and integrate beneficial external knowledge from multilingual collections. |
| Approach: | They propose a multilingual search-augmented reinforcement learning framework that integrates a language-coupled Group Relative Policy Optimization into the policy and reward models. |
| Outcome: | The proposed framework achieves competitive performance and is appropriate for various practical scenarios such as constrained training data and retrieval over collections encompassing a large number of languages. |
Get Confused Cautiously: Textual Sequence Memorization Erasure with Selective Entropy Maximization (2025.coling-main)
Copied to clipboard
| Challenge: | Existing methods for erasure of memorized text fail to unlearn large numbers of memorizable samples without jeopardizing model utility. |
| Approach: | They propose a method that allows LLMs to memorize and recite some training sequences verbatim . they propose an entropy-based loss method that is shown to be more stable . |
| Outcome: | The proposed method improves model utility and accuracy while preserving model ability in language generation and understanding. |
AlphaEdit+: Model Editing in the Presence of Conflicting and Inconsistent Knowledge (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for knowledge editing struggle with knowledge conflicts and inconsistencies. |
| Approach: | They propose a new method for knowledge editing that relaxes null-space constraints and introduces a weighting scheme to mitigate conflicts between new and historical knowledge. |
| Outcome: | The proposed method outperforms existing methods on challenging datasets and outperformed existing methods. |
Textual Dataset Distillation via Language Model Embedding (2024.findings-emnlp)
Copied to clipboard
| Challenge: | prevailing methods for dataset distillation generate distilled data as embedding vectors, which are not human-readable. |
| Approach: | They propose a model-agnostic, data-efficient method that leverages Language Model embeddings . their method offers enhanced flexibility and improved transferability . |
| Outcome: | The proposed method achieves comparable performance with faster processing times compared to other methods . it offers enhanced flexibility and improved transferability, expanding the range of potential applications . |
Margin Matching Preference Optimization: Enhanced Model Alignment with Granular Feedback (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for large language models rely on binary labels that fail to capture the subtle differences in relative quality between pairs. |
| Approach: | They propose a method that incorporates relative quality margins into optimization to improve LLM policies and reward models. |
| Outcome: | The proposed approach outperforms baseline methods on popular benchmarks including MT-bench and RewardBench. |
ReSo: A Reward-driven Self-organizing LLM-based Multi-Agent System for Reasoning Tasks (2025.emnlp-main)
Copied to clipboard
| Challenge: | Multi-agent systems (MAS) are limited by poor flexibility and scalability, with underdeveloped optimization strategies. |
| Approach: | They propose a task graph generation and a reward-driven two-stage agent selection process to integrate multi-agent systems to improve their reasoning capabilities. |
| Outcome: | The proposed model outperforms existing methods on Math-MAS and SciBench-MAS SciBech, while other methods completely fail. |
OmniDiagram: Advancing Unified Diagram Code Generation via Visual Interrogation Reward (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing studies on programmable diagram generation focus on a narrow set of tasks and languages. |
| Approach: | They propose a unified framework that integrates diverse diagram code languages and task definitions. |
| Outcome: | The proposed framework can bridge complex visual information with executable code across diverse tasks and languages. |
ProActor: Timing-Aware Reinforcement Learning for Proactive Task Scheduling Agents (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to measuring and optimizing proactive task-oriented agents lack generalizable end-to-end solutions. |
| Approach: | They propose a framework for conversational task scheduling that integrates proactiveness reinforcement learning with a domain-agnostic annotation methodology. |
| Outcome: | The proposed framework enables scalable proactiveness reinforcement learning (RL) Experiments on two newly auto-annotated datasets demonstrate significant improvements in proactive timing while maintaining action consistency comparable to state-of-the-art baselines. |
MoE Adapter for Large Audio Language Models: Sparsity, Disentanglement, and Gradient-Conflict-Free (2026.findings-acl)
Copied to clipboard
Yishu Lei, Shuwei He, Hu Jing, Dan Zhang, Xianlong Luo, Danxiang Zhu, Shikun Feng, Rui Liu, Jingzhou HE, Yu Sun, Hua Wu, Haifeng Wang
| Challenge: | Existing research on Large Language Models (LLMs) limited to textual input modality . acoustic information is intrinsically heterogeneous, entangling attributes such as speech, music, and environmental context. |
| Approach: | They propose a sparse Mixture-of-Experts architecture to decouple acoustic information by routing audio tokens to specialized experts. |
| Outcome: | The proposed architecture outperforms existing models on audio semantic and paralinguistic tasks while retaining shared experts for global context. |
QaRL: Rollout-Aligned Quantization-Aware RL for Fast and Stable Training under Training–Inference Mismatch (2026.findings-acl)
Copied to clipboard
Hao Gu, Hao Wang, Jiacheng Liu, Lujun Li, Qiyuan Zhu, Bei Liu, Binxing Xu, Lei Wang, Xintong Yang, Sida Lin, Sirui Han, Yike Guo
| Challenge: | Recent work has shown that reinforcement learning with simple rule-based reward functions (RLVR) can induce emergent reasoning behaviors and yield gains in challenging domains such as math problem solving. |
| Approach: | They propose a rollout-alignment-quantization-aware RL which aligns training-side forward with the quantized rollout to minimize mismatch. |
| Outcome: | The proposed approach outperforms quantized-rollout training by +5.5 on Qwen3-30B-A3B MoE for math problems while maintaining low-bit throughput. |
Team-Based Self-Play With Dual Adaptive Weighting for Fine-Tuning LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Recent self-training approaches have reduced reliance on human-labeled data, which limits their scalability. |
| Approach: | They propose a team-based self-play algorithm that iteratively refines alignment without additional human supervision. |
| Outcome: | The proposed algorithm outperforms baselines and LLM benchmarks in the self-supervised setting. |
PEGRL: Improving Machine Translation by Post-Editing Guided Reinforcement Learning (2026.findings-acl)
Copied to clipboard
| Challenge: | Reinforcement learning (RL) has shown strong promise for LLM-based machine translation . however, translation-oriented RL remains challenged by high-variance policy gradients induced by Monte Carlo baselines and large trajectory space that favors global exploration over fine-grained local optimization. |
| Approach: | They propose a two-stage RL framework that uses post-editing as an auxiliary task to stabilize training and guide overall optimization. |
| Outcome: | The proposed framework supports global exploration and fine-grained optimization while supporting global exploration. |
RanLoRA: Residual-aware Nonlinear Low-Rank Adaptation (2026.findings-acl)
Copied to clipboard
| Challenge: | Low-Rank Adaptation (LoRA) relying on linear low-rank projections restricts adaptation to linear subspaces, limiting flexibility on complex downstream tasks. |
| Approach: | They propose a nonlinear low-rank Adaptation approach that leverages pretrained weights to decompose them into principal components that are kept frozen and residual components that can be used for task-specific adaptation. |
| Outcome: | The proposed approach outperforms vanilla LoRA and representative variants on commonsense reasoning, image classification, and mathematical reasoning tasks. |
Context-attended Adversarial Reinforcement Learning for Robust Multi-step Retrieval Augmented Generation (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to multi-step retrieval-augmented generation are susceptible to retrieval noise and fabricated documents in real-world scenarios. |
| Approach: | They propose a framework for multi-step retrieval-augmented generation that incorporates external knowledge into a retriever to generate responses from adversarial samples. |
| Outcome: | The proposed framework improves performance in multiple noisy scenarios and can be used to improve multi-step retrieval-augmented generation. |
FocalOrder: Focal Preference Optimization for Reading Order Detection (2026.acl-long)
Copied to clipboard
Fuyuan Liu, Dianyu Yu, He Ren, Nayu Liu, Xiaomian Kang, Delai Qiu, Fa Zhang, Genpeng Zhen, Shengping Liu, Liang Jiaen, null Weihuang, Yining Wang, Junnan Zhu
| Challenge: | Existing methods for document comprehension rely on uniform supervision, resulting in a performance degradation in the intermediate sections. |
| Approach: | They propose a framework driven by Focal Preference Optimization to detect reading order in document layouts. |
| Outcome: | The proposed framework outperforms competing baselines and surpasses large-scale general VLMs. |
Survival of the Most Influential Prompts: Efficient Black-Box Prompt Search via Clustering and Pruning (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Prompt-based learning has been an effective paradigm for large pretrained language models (LLMs), enabling few-shot or even zero-shot learning. |
| Approach: | They propose a black-box prompt search method that clusters and prunes the search space to focus exclusively on influential prompt tokens. |
| Outcome: | The proposed method achieves state-of-the-art performance across tasks and LLMs while significantly reducing search costs. |
FastDiSS: Few-step Match Many-step Diffusion Language Model on Sequence-to-Sequence Generation (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing models that correct errors in the model but lack a high quality of output . a novel training framework that matches inference noise to the model's inference signal improves performance . |
| Approach: | They propose a training framework that matches inference noise to model errors by perturbing the self-conditioning signal to match inference errors. |
| Outcome: | The proposed framework surpasses standard continuous diffusion models while providing 400x faster inference speed. |
Two-Stage Fine-Tuning for Improved Bias and Variance for Large Pretrained Language Models (2023.acl-long)
Copied to clipboard
| Challenge: | Recent work challenges the bias-variance trade-off . large pretrained models can have large variance and overfit domain-specific data . |
| Approach: | They propose a bias-variance trade-off that implies learning methods need to balance complexity with data size to minimize under-fitting and over-fit. |
| Outcome: | The proposed method achieves strong results on SuperGLUE and clinical information extraction tasks. |
RAG-RewardBench: Benchmarking Reward Models in Retrieval Augmented Generation for Preference Alignment (2025.findings-acl)
Copied to clipboard
Zhuoran Jin, Hongbang Yuan, Tianyi Men, Pengfei Cao, Yubo Chen, Jiexin Xu, Huaijun Li, Xiaojian Jiang, Kang Liu, Jun Zhao
| Challenge: | Existing retrieval augmented language models often overlook effective alignment with human preferences. |
| Approach: | They propose a benchmark to evaluate RMs in retrieval augmented language models . they incorporate 18 RAG subsets, six retrievers, and 24 RALMs to increase diversity . |
| Outcome: | The proposed benchmark combines 18 RAG subsets, six retrievers, and 24 RALMs to increase diversity of data sources. |
Maximizing Local Entropy Where It Matters: Prefix-Aware Localized LLM Unlearning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to machine unlearning treat all tokens indiscriminately and enforce uncertainty over the entire vocabulary. |
| Approach: | They propose a framework that targets the prefix in a response and minimizes uncertainty in the critical subspace. |
| Outcome: | The proposed framework achieves superior forgetting efficacy and utility preservation compared to baselines. |
SeaPO: Strategic Error Amplification for Robust Preference Optimization of Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for preference optimization of large language models use pairs of positive and negative samples, but the quality of positive samples may become similar during training, complicating preference learning. |
| Approach: | SeaPO introduces error types commonly occurring in large language models to improve preference learning. |
| Outcome: | SeaPO introduces error types into model Preference Optimization to improve model performance . negative samples are more erroneous than positive samples, and preference-based training mitigates errors . |
Beyond Uniform SVD: Dual-Level Optimization across Columns and Modules for LLM Compression (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for low-rank decomposition overlook decomposing errors and suboptimal approximation. |
| Approach: | They propose a low-rank decomposition framework that integrates low-level optimization at column and module levels. |
| Outcome: | The proposed framework outperforms state-of-the-art methods and baselines in SVD and pruning. |
Preconditioned Test-Time Adaptation for Out-of-Distribution Debiasing in Narrative Generation (2026.acl-long)
Copied to clipboard
| Challenge: | Debiased large language models excel at handling known or low-bias prompts, but fail on unfamiliar and high-biased prompts. |
| Approach: | They propose a debiasing framework that detects high-bias prompts and triggers context-aware LoRA updates only when a bias-risk score exceeds a threshold. |
| Outcome: | The proposed framework reduces toxicity/bias score with significantly lower latency than standard optimization methods. |
OMS: On-the-fly, Multi-Objective, Self-Reflective Ad Keyword Generation via LLM Agent (2025.emnlp-main)
Copied to clipboard
| Challenge: | Keyword decision in Sponsored Search Advertising is critical to the success of ad campaigns. |
| Approach: | They propose a keyword generation framework that is On-the-fly and Multi-objective to automate keyword generation. |
| Outcome: | Experiments show that OMS outperforms existing methods in keyword generation . relying on large-scale query-keyword data is a major limitation, authors say . |
No perspective, no perception!! Perspective-aware Healthcare Answer Summarization (2024.findings-acl)
Copied to clipboard
| Challenge: | Healthcare Community Question Answering forums are prone to off-topic discussions and diverse answers can be challenging for readers to sift through. |
| Approach: | They propose a task of perspective-specific answer summarization to identify different perspectives within healthcare-related responses and frame a perspective-driven abstractive summary covering all responses. |
| Outcome: | The proposed model outperforms existing models against five baselines and shows that it is more accurate than existing models. |
Cramming 1568 Tokens into a Single Vector and Back Again: Exploring the Limits of Embedding Space Capacity (2025.acl-long)
Copied to clipboard
| Challenge: | Recent work addresses problem of compression of tokens into shorter sequence of real-valued vectors . attainable lossless compression ratio is typically not higher than x10 . |
| Approach: | They propose to compress a sequence of tokens into a shorter sequence of real-valued vectors to be used as inputs instead of token embeddings or key-value cache. |
| Outcome: | The proposed algorithms reduce the amount of compute in existing language models rather than minimizing number of bits needed to store text. |
HyperLoRA: Efficient Cross-task Generalization via Constrained Low-Rank Adapters Generation (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to adapt pre-trained language models (PLMs) to emerging tasks are costly and inefficient. |
| Approach: | They propose a meta-network that generates task-specific weights without any optimization. |
| Outcome: | The proposed approach has flexible generalization ability and superior performance over hypenetworks. |
Debate as Optimization: Adaptive Conformal Prediction and Diverse Retrieval for Event Extraction (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Experimental results show a significant performance gap between tuning-based approaches and event extraction approaches. |
| Approach: | They propose a debate as optimization system where the primary objective is to iteratively refine the large language models outputs through debating without parameter tuning. |
| Outcome: | The proposed system reduces performance gap between supervised approaches and tuning-free methods by 18.1% and 17.8% on ACE05 and 17.9% and 15.2% on CASIE respectively. |
IAM: Efficient Inference through Attention Mapping between Different-scale LLMs (2025.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) are a challenge due to their internal reasoning processes. |
| Approach: | They propose an algorithm that can optimize attention matrices by performing attention mapping between small and large LLMs. |
| Outcome: | The proposed framework can reduce KV cache usage by 22.1% and accelerate prefill by 15% without sacrificing performance. |
ACING: Actor-Critic for Instruction Learning in Black-Box LLMs (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have impressive capabilities across tasks like classification, summarization, and reasoning. |
| Approach: | They propose an actor-critic reinforcement learning framework that formulates instruction optimization as a stateless, continuous-action problem. |
| Outcome: | The proposed framework outperforms human-written prompts in 76% of instruction-induction tasks with gains of 33 points and 10-point improvement over baseline. |
MA-GTS: A Multi-Agent Framework for Solving Complex Graph Problems in Real-World Applications (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for solving complex problems are expensive and inefficient when handling large-scale, high-complexity problems. |
| Approach: | They propose a multi-agent framework that decomposes complex problems through agent collaboration by mapping implicitly expressed graph data into clear, structured graph representations and dynamically selecting the most suitable algorithm based on problem constraints and graph structure scale. |
| Outcome: | The proposed framework outperforms state-of-the-art methods on multiple benchmarks with robust performance on both closed- and open-source models. |
Specialization without Sparsity: Efficient and Expressive Split-Path Experts for LLM Fine-Tuning (2026.findings-acl)
Copied to clipboard
| Challenge: | Parameter-efficient fine-tuning (PEFT) is a low-cost alternative to full fine-timing due to the massive overhead. |
| Approach: | They propose a Mixture-of-Experts approach that enhances specialization while maintaining low resource overhead. |
| Outcome: | The proposed approach outperforms or matches state-of-the-art methods on GLUE, GSM8K, MBPP, and a text rewriting task from SmolTalk. |
S2ST-Omni: Hierarchical Language-Aware SpeechLLM Adaptation for Multilingual Speech-to-Speech Translation (2026.findings-acl)
Copied to clipboard
| Challenge: | S2ST-Omni integrates a speech-to-text frontend with a modular, plug-and-play text-tospeech backend. |
| Approach: | They propose a compositional S2ST framework that integrates a speech-to-text frontend with a modular, plug-and-play text-tospeech backend. |
| Outcome: | The proposed framework outperforms existing frameworks in translation and synthesis . it integrates a speech-to-text translation frontend with a plug-and-play text-tospeech backend . |
Counteracting the Matthew Effect in Self-Improvement of LVLMs through Head-Tail Re-balancing (2026.acl-long)
Copied to clipboard
Xin Guo, Zhiheng Xi, Yiwen Ding, Yitao Zhai, Xiaowei Shi, Xunliang Cai, Tao Gui, Qi Zhang, Xuanjing Huang
| Challenge: | Large vision language models have impressive reasoning capabilities across complex multimodal tasks. |
| Approach: | They propose to use distribution-reshaping and trajectory-rebalancing to improve visual reasoning capabilities. |
| Outcome: | Experiments on Qwen2-VL-7B-Instruct and InternVL2.5-4B models show that their methods outperform baselines by 3.86 points. |
How Do LLMs Acquire New Knowledge? A Knowledge Circuits Perspective on Continual Pre-Training (2025.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have exceptional capabilities in knowledge-intensive tasks . however, they struggle with knowledge updates due to dynamic nature of world knowledge . |
| Approach: | They propose to identify computational subgraphs that facilitate knowledge storage and processing . they also identify a phase shift from formation to optimization in LLMs . |
| Outcome: | The proposed model can capture factual knowledge from pre-training corpus and encapsulate it as extensive parametric knowledge. |
Composable Text Controls in Latent Space with ODEs (2023.emnlp-main)
Copied to clipboard
Guangyi Liu, Zeyu Feng, Yuan Gao, Zichao Yang, Xiaodan Liang, Junwei Bao, Xiaodong He, Shuguang Cui, Zhen Li, Zhiting Hu
| Challenge: | Existing approaches to composable text operations often require plug-and-play . a single LM can perform arbitrary text operation composition in the latent space . |
| Approach: | They propose an efficient approach for composable text operations in the latent space of text . they connect pretrained LMs to the laten space and adapt them to the space . |
| Outcome: | The proposed approach improves on existing methods in the latent space of text. |
ClueAnchor: Clue-Anchored Knowledge Reasoning Exploration and Optimization for Retrieval-Augmented Generation (2025.findings-emnlp)
Copied to clipboard
Hao Chen, Yukun Yan, Sen Mei, Wanxiang Che, Zhenghao Liu, Qi Shi, Xinze Li, Yuchun Fan, Pengcheng Huang, Qiushi Xiong, Zhiyuan Liu, Maosong Sun
| Challenge: | Existing RAG systems often underutilize the retrieved documents, authors say . they fail to extract and integrate key clues needed to support faithful and interpretable reasoning . |
| Approach: | a new framework extracts key clues from retrieved content and generates multiple reasoning paths . the framework optimizes the model by selecting the most appropriate reasoning path . |
| Outcome: | Experiments show that ClueAnchor outperforms baseline RAG frameworks in completeness and robustness. |
MWPO: Enhancing LLMs Performance through Multi-Weight Preference Strength and Length Optimization (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing offline alternatives to Reinforcement Learning from Human Feedback (RLHF) are available at https://github.com/AIR-hl/MWPO. |
| Approach: | They propose an offline method to optimize preference pairs based on implicit reward margins and response length margins by reweighting them using a geometric mixture. |
| Outcome: | The proposed method outperforms state-of-the-art methods on four different scales and reduces generation length by 9.4%. |
DEM: Distribution Edited Model for Training with Mixed Data Distributions (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent fine-tuning approaches for large language models require supervised finetun on diverse datasets and follow different distributions. |
| Approach: | They propose a distribution edited model that integrates models individually trained on each data source with the base model using basic element-wise vector operations. |
| Outcome: | The proposed model outperforms baseline models on a variety of benchmarks and is cheaper than standard data mixing methods. |
CoTD-PO: Chain-of-Thought Distillation with Preference Optimization (2025.findings-emnlp)
Copied to clipboard
Lujie Niu, Haochen Sun, Fangkun Zhao, Sheng Chen, Zimeng Bai, Jiawei Zhang, Caixia Yuan, Xiaojie Wang
| Challenge: | Existing methods for chain-of-thought distillation suffer from a distribution mismatch between teacher-generated training trajectories and the student model's own generative distribution. |
| Approach: | They propose a framework that shifts the training paradigm from passive imitation to active trajectory exploration by allowing students to sample their own answer paths. |
| Outcome: | The proposed method outperforms standard CoT distillation baselines while mitigating mode collapse and preserving semantic diversity. |
Understanding and Leveraging the Expert Specialization of Context Faithfulness in Mixture-of-Experts LLMs (2025.emnlp-main)
Copied to clipboard
| Challenge: | Context faithfulness is essential for reliable reasoning in context-dependent scenarios. |
| Approach: | They propose a method that identifies and fine-tunes context-faithful experts . they propose 'context-faither fine- tuning' which selectively fine- tunes them . |
| Outcome: | The proposed method identifies experts with specialization in context utilization and improves context grounding. |
Heuristic-based Search Algorithm in Automatic Instruction-focused Prompt Optimization: A Survey (2025.findings-acl)
Copied to clipboard
Wendi Cui, Jiaxin Zhang, Zhuohang Li, Hao Sun, Damien Lopez, Kamalika Das, Bradley A. Malin, Sricharan Kumar
| Challenge: | Recent advances in Large Language Models (LLMs) have led to remarkable achievements across a variety of NLP tasks. |
| Approach: | They propose a taxonomy of automatic prompt optimization methods that explore and improve prompts with minimal human oversight. |
| Outcome: | The proposed methods can explore and improve prompts with minimal human oversight. |
CSPO: Alleviating Reward Ambiguity for Structured Table-to-LaTeX Generation (2026.acl-long)
Copied to clipboard
| Challenge: | Tables contain rich structured information, but when stored as images their contents remain "locked" within pixels. |
| Approach: | They propose a framework that disentangles optimization across LaTeX tables components . CSPO assigns component-specific rewards and backpropagates each signal through tokens . |
| Outcome: | The proposed framework disentangles optimization across LaTeX tables components—structure, style, and content. |
Task Facet Learning: A Structured Approach To Prompt Optimization (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to prompt optimization are limited to learning multiple facets of a task from training examples. |
| Approach: | They propose to optimize a text prompt by considering different facets of a task and including them in the prompt. |
| Outcome: | The proposed algorithm can generate long, complex prompts that existing methods are unable to generate. |
Golden Touchstone: A Comprehensive Bilingual Benchmark for Evaluating Financial Large Language Models (2025.findings-emnlp)
Copied to clipboard
Xiaojun Wu, Junxi Liu, Huan-Yi Su, Zhouchi Lin, Yiyan Qi, Chengjin Xu, Jiajun Su, Jiajie Zhong, Fuwei Wang, Saizhuo Wang, Fengrui Hua, Jia Li, Jian Guo
| Challenge: | Existing financial benchmarks suffer from limited language and task coverage, low-quality datasets, and inadequate adaptability for LLM evaluation. |
| Approach: | They propose a bilingual benchmark for financial LLMs that assesses models’ language understanding and generation capabilities. |
| Outcome: | The proposed bilingual benchmark assesses models’ language understanding and generation capabilities. |
What Makes an LLM a Good Optimizer? A Trajectory Analysis of LLM-Guided Evolutionary Search (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent work has demonstrated the promise of orchestrating large language models (LLMs) within evolutionary and agentic optimization systems. |
| Approach: | They present a large-scale study of LLM-guided evolutionary search . they find strong LLMs behave as local refiners, producing frequent improvements . weaker LLM optimizers exhibit large semantic drift, they say . |
| Outcome: | The results highlight the importance of trajectory analysis for understanding and improving LLM-based optimization systems. |
Conflict-Aware Memory for Embodied Agents: Enhancing Vector Data Quality via Detection Rules (2026.acl-long)
Copied to clipboard
| Challenge: | Embodied agents have successfully leveraged large language models (LLMs) to better transform human instructions and images into executable task plans. |
| Approach: | They propose Conflict Detection Rules to identify and manage data quality issues in vector knowledge bases and correct the index structure. |
| Outcome: | Experimental results show that planners with Conflict Detection Rules exceed the basic LLM planner by 15.25% and 14.25% in grammatical accuracy (GA) and interpretation accuracy (IA) on average. |
OASIS: Mitigating Harmful Fine-tuning Attacks on LLMs via Orthogonal and Adaptive Safety Alignment Strategy (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods to decouple safety enforcement from harmful feature acquisition rely on perturbation directions that conflict with harmful gradients . harmful fine-tuning attacks pose a significant challenge for service providers aiming to uphold rigorous safety standards. |
| Approach: | They propose an orthogonal and ad hoc safety alignment strategy to decouple safety enforcement from harmful feature acquisition. |
| Outcome: | Experiments on four large language models show that OASIS reduces the Harmful Score by 60% compared to baselines while maintaining stable task utility. |
AgentDrug: Utilizing Large Language Models in an Agentic Workflow for Zero-Shot Molecular Optimization (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Molecular optimization is a fundamental task in drug discovery. |
| Approach: | They propose an agentic workflow that leverages LLMs in a structured refinement process to achieve significantly higher accuracy. |
| Outcome: | The proposed workflow improves on single- and multi-property optimization tasks under loose and strict thresholds. |
T-REG: Preference Optimization with Token-Level Reward Regularization (2025.acl-long)
Copied to clipboard
| Challenge: | Reinforcement learning from human feedback (RLHF) is a dominant approach for large language models to follow instructions and produce meaningful alignment. |
| Approach: | They propose a method that leverages human feedback to optimize large language models . they propose to use sequence-level and token-level rewards to optimize preference . |
| Outcome: | The proposed method outperforms baseline methods on Alpaca Eval 2 and Arena-Hard benchmarks. |
Unifying Uniform and Binary-coding Quantization for Accurate Compression of Large Language Models (2025.acl-long)
Copied to clipboard
Seungcheol Park, Jeongin Bae, Beomseok Kwon, Minjun Kim, Byeongwook Kim, Se Jung Kwon, U Kang, Dongsoo Lee
| Challenge: | Quantization is essential for deploying large language models (LLMs) efficiently since they require expensive computational and memory costs. |
| Approach: | They propose a quantization method that unifies flexible mapping techniques to optimize parameters precisely. |
| Outcome: | The proposed method outperforms existing methods and achieves higher accuracy on GSM8K benchmark. |
MeasHalu: Mitigation of Scientific Measurement Hallucinations for Large Language Models with Enhanced Reasoning (2026.findings-acl)
Copied to clipboard
Ruijun Huang, Zhiqiao Kang, Yuxuan Zhu, Junxiong Li, Jiahao Zhao, Minghuan Tan, Feng Jiang, Min Yang
| Challenge: | Large Language Models (LLMs) exhibit severe hallucinations, which undermine reliability of automated scientific document understanding systems. |
| Approach: | They propose a framework for mitigating scientific measurement hallucinations through enhanced reasoning and targeted optimization. |
| Outcome: | The proposed framework significantly reduces hallucination rates and improves overall accuracy on the MeasEval benchmark. |
End-to-End Optimization of LLM-Driven Multi-Agent Search Systems via Heterogeneous-Group-Based Reinforcement Learning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing multi-agent reinforcement learning methods depend on large critic networks to evaluate joint actions, leading to instability and high memory costs. |
| Approach: | They propose a method to optimize large language models for agent-specific roles . they propose combining agent-based frameworks with retrieval-augmented generation . |
| Outcome: | Experiments show that multi-agent group policy optimization outperforms baselines in task performance and computational efficiency. |
Reviving Cultural Heritage: A Novel Approach for Comprehensive Historical Document Restoration (2025.acl-long)
Copied to clipboard
Yuyi Zhang, Peirong Zhang, Zhenhua Yang, Pengyu Yan, Yongxin Shi, Pengwei Liu, Fengjun Guo, Lianwen Jin
| Challenge: | Existing methods for historical document restoration focus on single modality or limited-size restoration, failing to meet practical needs. |
| Approach: | They propose a full-page HDR dataset and an automated HDR solution to replace manual restoration methods. |
| Outcome: | The proposed solution improves OCR accuracy from 46.83% to 84.05% when processing severely damaged documents, with enhancement to 94.25% through human-machine collaboration. |
Forge: Quality-Aware Reinforcement Learning for NP-Hard Optimization in LLMs (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing benchmarks focus on correctness, overlooking optimality . large language models excel at math, coding, logic and puzzles . |
| Approach: | They propose a framework for training and evaluating Large Language Models on NP-hard optimization problems through quality-aware RLVR. |
| Outcome: | The proposed framework outperforms existing benchmarks on math, coding, logic and puzzles. |
Bypass Back-propagation: Optimization-based Structural Pruning for Large Language Models via Policy Gradient (2025.acl-long)
Copied to clipboard
| Challenge: | Recent pruning methods rely on heuristically hand-crafted metrics, leading to suboptimal performance. |
| Approach: | They propose a method that optimizes pruning masks by minimizing back-propagation . they learn an underlying Bernoulli distribution to sample binary pruning mask samples . |
| Outcome: | The proposed method is able to support global and heterogeneous pruning without back-propagation. |
GFT: From Imitation to Reward Fine-Tuning with Unbiased Group Advantages and Dynamic Coefficient Rectification (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing studies have demonstrated that supervised fine-tuning and reinforcement learning are effective in integrating knowledge injection with robust generalization. |
| Approach: | They propose a unified post-training framework that addresses intrinsic limitations of supervised fine-tuning and reinforcement learning. |
| Outcome: | The proposed framework surpasses SFT-based methods and yields policies that integrate more smoothly with subsequent RL training. |
Cross-Modal Masked Compositional Concept Modeling for Enhancing Visio-Linguistic Compositionality (2026.acl-long)
Copied to clipboard
| Challenge: | a contrastive learning approach for vision-language models is needed to capture compositional information. |
| Approach: | They propose a framework that masks compositional concepts in one modality and reconstructs them conditioned on full contextual information from the other . |
| Outcome: | The proposed framework enhances compositionality in visual language models and improves their ability to capture syntactic structure and linguistic information. |
ScaleBiO: Scalable Bilevel Optimization for LLM Data Reweighting (2025.acl-long)
Copied to clipboard
Rui Pan, Dylan Zhang, Hanning Zhang, Xingyuan Pan, Minrui Xu, Jipeng Zhang, Renjie Pi, Xiaoyu Wang, Tong Zhang
| Challenge: | Existing paradigms for bilevel optimization require second-order information, making it difficult to scale them up. |
| Approach: | They propose a scalable instantiation of a bilevel optimization paradigm for large-scale LLMs by using a memory-efficient training technique. |
| Outcome: | The proposed paradigm scales to 30B-sized LLMs on 8H100 GPUs. |
EducationQ: Evaluating LLMs’ Teaching Capabilities Through Multi-Agent Dialogue Framework (2025.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are increasingly used as educational tools, yet evaluating their teaching capabilities remains challenging due to the resource-intensive nature of teacher-student interactions. |
| Approach: | They propose a multi-agent dialogue framework that efficiently assesses teaching capabilities through simulated dynamic educational scenarios. |
| Outcome: | The proposed framework outperforms open-source models on 1,498 questions across 13 disciplines and 10 difficulty levels on 1,400 questions. |
NL2Lean: Translating Natural Language into Lean 4 through Multi-Aspect Reinforcement Learning (2025.emnlp-main)
Copied to clipboard
Yue Fang, Shaohan Huang, Xin Yu, Haizhen Huang, Zihan Zhang, Weiwei Deng, Furu Wei, Feng Sun, Qi Zhang, Zhi Jin
| Challenge: | Existing formal proof assistants rely on instruction tuning and lack fine-grained structural and semantic alignment. |
| Approach: | They propose a reinforcement learning framework that enables LLMs to translate natural language into formal language such as Lean 4 . they use a model with basic translation ability to refine the model's reinforcement learning . |
| Outcome: | The proposed method outperforms baseline models on NL-to-Lean 4 tasks. |
LLM-Powered Benchmark Factory: Reliable, Generic, and Efficient (2026.acl-long)
Copied to clipboard
Peiwen Yuan, Shaoxiong Feng, Yiwei Li, Xinglin Wang, Yueqi Zhang, Jiayi Shi, Chuyi Tan, Boyuan Pan, Yao Hu, Kan Li
| Challenge: | Using generic and efficient benchmark generators, human annotators are limited by inefficiency . current benchmark generator methods rely on seed signals, leading to long cycles and high costs . |
| Approach: | They propose a framework to evaluate LLMs as generic benchmark generators and integrate them as BenchMaker. |
| Outcome: | The proposed framework achieves comparable performance to human-annotated benchmarks on most metrics. |
Mind Reader: Latent User Demand-Guided Content Optimization for Generative Search Engine (2026.acl-long)
Copied to clipboard
Tong Chen, JiaWei Guo, Yuxi Li, Baiming Chen, Houxing Ren, Zhang Zhiwei, Yunxiang Zhang, Hanyang Xia, Kun Liang, Zhaoran Fan
| Challenge: | Generative Search Engines (GSEs) have reshaped information retrieval and Generating Engine Optimization (GEO) emerges to improve the content visibility in GSEs’ responses. |
| Approach: | They propose a method to optimize content to cover latent semantic information of GSEs by decomposing query into diverse perspectives and capturing underlying semantic information. |
| Outcome: | The proposed method outperforms baselines and effectively improves content visibility (with up to 2.44x objective metrics and 1.23x subjective metrics on average). |
ProMCP: Profiling Token Flows and Latency Costs in Model Context Protocol–Based LLM Agents (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models are increasingly used as agents that interact with external tools and data sources to solve tasks that require fresh knowledge, precise computation, or action in a real environment. |
| Approach: | They propose a framework that decomposes the MCP workflow into a six-stage communication pipeline and enables granular attribution of computational costs. |
| Outcome: | The proposed framework decomposes the MCP workflow into a six-stage communication pipeline, enabling granular attribution of computational costs. |
MolMem: Memory-Augmented Agentic Reinforcement Learning for Sample-Efficient Molecular Optimization (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for molecular optimization use expensive oracle evaluations to achieve sample efficiency under limited oracular budget. |
| Approach: | They propose a framework that iteratively refines a lead compound to improve molecular properties while preserving structural similarity to the original molecule. |
| Outcome: | The proposed framework achieves 90% success on single-property tasks and 52% on multi-propety task using only 500 oracle calls. |
Beyond Fully Random Masking: Attention-Guided Denoising and Optimization for Diffusion Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for full-attention dLLMs rely on random masking strategies that overlook intrinsic token dependencies. |
| Approach: | They propose an attention-guided denoising and optimization framework that aligns training and optimization with attention-derived dependencies. |
| Outcome: | The proposed framework outperforms state-of-the-art methods on mathematical and coding benchmarks. |
VL-Calibration: Decoupled Confidence Calibration for Large Vision-Language Models Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing verbalized confidence calibration methods for large vision language models optimize a single holistic confidence score using binary answer-level correctness. |
| Approach: | They propose a reinforcement learning framework that explicitly decouples confidence into visual and reasoning confidence. |
| Outcome: | Experiments show that the proposed framework decouples confidence into visual and reasoning confidence while suppressing ungrounded hallucinations while preserving valid perception. |
Prompt Optimization for Relation Extraction using Reinforcement Learning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing prompt-based methods rely heavily on large-scale annotated datasets limiting their applicability in domain-specific and low-resource scenarios. |
| Approach: | They propose a reinforcement learning-based automated prompt optimization framework for domain relation extraction that optimizes prompt quality through interaction with a black-box LLM. |
| Outcome: | The proposed framework outperforms existing prompt-based methods and supervised baselines on multiple extraction datasets across medical, financial, legal, and news domains. |
Multi-Task Representation Alignment on Language Understanding: A Mutual Information Perspective (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to multitask learning fail to address task interference issues . Existing methods focus on task balancing or probabilistic modeling but fail to learn sufficient representations for all target tasks. |
| Approach: | They propose a multi-task representation alignment framework to achieve task-specific alignment and self-alignment on shared representations from a mutual information perspective. |
| Outcome: | The proposed framework outperforms 13 representative MTL methods under label-noisy and data-constrained conditions. |
Reinforcement Learning for Diffusion LLMs via Energy-Based Gibbs Alignment (2026.acl-long)
Copied to clipboard
| Challenge: | Diffusion Large Language Models (dLLMs) offer parallel decoding and bidirectional context modeling . aligning dLLms with reinforcement learning (RL) remains a challenge . |
| Approach: | They propose a variational framework that reformulates RL for dLLMs as a distribution matching problem. |
| Outcome: | The proposed framework reformulates RL for dLLMs as a distribution matching problem. |