Papers by Shi Feng

SwiftPrune: Hessian-Free Weight Pruning for Large Language Models (2025.findings-emnlp)

Challenge:	Existing research on multimodal dialogues focuses on textual response generation and visual response selection based on the dialogue context.
Approach:	They propose a generative model framework for multimodal dialogue response generation that ground the conversation on an image.
Outcome:	The proposed system provides users with an enhanced conversational experience.

Copied to clipboard

Yuhan Kang, Yang Shi, Mei Wen, Jun He, Jianchao Yang, Zeyu Xue, Jing Feng, Xinwang Liu

Empowering Math Problem Generation and Reasoning for Large Language Model via Synthetic Data based Continual Learning Framework (2025.emnlp-main)

Challenge:	a novel post-training pruning method relies on the Hessian matrix to perform pruning . current pruning methods are computationally intensive and lack performance due to second-order derivative calculations.
Approach:	They propose a Hessian-free weight pruning method that reduces computational burden . they use an Exponentially Weighted Moving Average technique to bypass weight sorting .
Outcome:	The proposed method achieves hardware-efficient model compression by eliminating computational intensive calculations.

Copied to clipboard

Qian Wan, Wangzi Shi, Jintian Feng, Shengyingjie Liu, Luona Wei, Zhicheng Dai, Jianwen Sun

Direct Multi-Turn Preference Optimization for Language Agents (2024.emnlp-main)

Challenge:	Existing learning frameworks for large language models (LLMs) for math problem generation are limited and lack quality data.
Approach:	They propose a synthetic data based continual learning framework to improve LLMs ability for MPG and math reasoning.
Outcome:	The proposed framework improves performance on large language models and math reasoning using supervised fine-tuning, data synthesis and direct preference optimization.

Copied to clipboard

Wentao Shi, Mengqi Yuan, Junkang Wu, Qifan Wang, Fuli Feng

Challenge:	Extensive experiments on three multi-turn agent task datasets confirm the effectiveness and superiority of the DMPO loss function.
Approach:	They propose a novel loss function for multi-turn agent tasks that replaces the policy constraint with the state-action occupancy measure constraint and adds length normalization to the Bradley-Terry model.
Outcome:	Experiments on three multi-turn agent task datasets confirm the effectiveness and superiority of the proposed loss function.

STICKERCONV: Generating Multimodal Empathetic Responses from Scratch (2024.acl-long)

Copied to clipboard

Yiqun Zhang, Fanheng Kong, Peidong Wang, Shuang Sun, SWangLing SWangLing, Shi Feng, Daling Wang, Yifei Zhang, Kaisong Song

Resource-Limited Joint Multimodal Sentiment Reasoning and Classification via Chain-of-Thought Enhancement and Distillation (2026.findings-acl)

Challenge:	Prior studies on stickers focused on sentiment analysis and recommendation systems, overlooking their vast potential in empathetic response generation.
Approach:	They propose a multimodal empathetic dialogue dataset, STICKERCONV, which simulates human behavior with stickers, and propose evaluative metrics based on LLM.
Outcome:	The proposed framework generates contextually relevant and emotionally resonant multimodal empathetic responses, contributing to the advancement of more nuanced and engaging e-dialog systems.

Copied to clipboard

Haonan Shangguan, Xiaocui Yang, Shi Feng, Daling Wang, Yifei Zhang, Feiliang Ren, Ge Yu

Information Retrieval

Challenge:	Current approaches for Multimodal Sentiment Analysis (MSA) rely on parameter-heavy LLMs for classification, overlooking multimodal sentiment reasoning generation in resource-limited environments.
Approach:	They propose a multimodal sentiment reasoning distillation model that employs a teacher-assistant-student paradigm to address deployment constraints in resource-limited environments.
Outcome:	The proposed model performs well on a resource-limited JMSRC task with only 3B parameters and shows generalization and interpretability.

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

Universal Adversarial Triggers for Attacking and Analyzing NLP (D19-1)

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.

Copied to clipboard

Eric Wallace, Shi Feng, Nikhil Kandpal, Matt Gardner, Sameer Singh

Personalized Microblog Sentiment Classification via Adversarial Cross-lingual Multi-task Learning (D18-1)

Challenge:	Using adversarial triggers, a model can produce a specific prediction . adversarial attacks are useful for evaluation and interpretation .
Approach:	They propose a gradient-guided search over tokens that finds short adversarial triggers that successfully trigger the target prediction.
Outcome:	The proposed algorithm finds short trigger sequences that successfully trigger the target prediction.

Copied to clipboard

Weichao Wang, Shi Feng, Wei Gao, Daling Wang, Yifei Zhang

SEGMENT+: Long Text Processing with Short-Context Language Models (2024.emnlp-main)

Challenge:	Existing personalized microblog sentiment classification methods suffer from the insufficiency of discriminative tweets for personalization learning.
Approach:	They propose to use user-attention-based Convolutional Neural Networks to capture individuality and opinion bias in microblog posts and a novel adversarial cross-lingual learning framework to enrich the user post representation.
Outcome:	The proposed method outperforms state-of-the-art baseline algorithms with large margins on English and Chinese microblog datasets.

Copied to clipboard

Wei Shi, Shuang Li, Kerun Yu, Jinglei Chen, Zujie Liang, Xinhui Wu, Yuxi Qian, Feng Wei, Bo Zheng, Jiaqing Liang, Jiangjie Chen, Yanghua Xiao

MulZDG: Multilingual Code-Switching Framework for Zero-shot Dialogue Generation (2022.coling-1)

Challenge:	Existing frameworks that increase context window do not guarantee robust performance across long input tasks.
Approach:	They propose a framework that enables language models to handle extended inputs within limited context windows efficiently.
Outcome:	The framework improves performance on long-document question-answering and Needle-in-a-Haystack tasks.

Copied to clipboard

Yongkang Liu, Shi Feng, Daling Wang, Yifei Zhang

TexSmart: A System for Enhanced Natural Language Understanding (2021.acl-demo)

Challenge:	Existing zero-shot dialogue generation systems rely on large-scale pre-trained language models.
Approach:	They propose a multilingual learning framework for zero-shot dialogue generation that can transfer knowledge from an English corpus to a non-English corpus with zero samples.
Outcome:	The proposed framework can transfer knowledge from an English corpus to a non-English corpus with zero samples.

Copied to clipboard

Lemao Liu, Haisong Zhang, Haiyun Jiang, Yangming Li, Enbo Zhao, Kun Xu, Linfeng Song, Suncong Zheng, Botong Zhou, Dick Zhu, Xiao Feng, Tao Chen, Tao Yang, Dong Yu, Feng Zhang, ZhanHui Kang, Shuming Shi

Information Extraction & Text Mining

Challenge:	TexSmart supports fine-grained named entity recognition (NER) Large-scale fine-granular entity types are expected to provide richer semantic information for downstream NLP applications.
Approach:	They introduce TexSmart, a text understanding system that supports fine-grained named entity recognition (NER) and enhanced semantic analysis functionalities.
Outcome:	The proposed system supports fine-grained named entity recognition (NER) and enhanced semantic analysis functions.

TS-Align: A Teacher-Student Collaborative Framework for Scalable Iterative Finetuning of Large Language Models (2024.findings-emnlp)

Copied to clipboard

Chen Zhang, Chengguang Tang, Dading Chong, Ke Shi, Guohua Tang, Feng Jiang, Haizhou Li

GAPO: Learning Preferential Prompt through Generative Adversarial Policy Optimization (2025.acl-long)

Challenge:	Mainstream approaches to aligning large language models heavily rely on human preference data.
Approach:	They propose a framework that fine-tunes a policy model using pairwise feedback data automatically mined from its outputs.
Outcome:	The proposed framework outperforms the base model with an average win rate of 69.7% across seven conversational or instruction-following datasets.

Copied to clipboard

Zhouhong Gu, Xingzhou Chen, Xiaoran Shi, Tao Wang, Suhang Zheng, Tianyu Li, Hongwei Feng, Yanghua Xiao

ToolBeHonest: A Multi-level Hallucination Diagnostic Benchmark for Tool-Augmented Large Language Models (2024.emnlp-main)

Challenge:	Existing methods for achieving this require a limited understanding of constraints and can be hallucinating or brittle.
Approach:	They propose a framework that combines adversarial training dynamics with an encoder-only reward model to progressively learn and adapt to increasingly complex constraints.
Outcome:	Extensive experiments show that GAPO significantly outperforms existing methods like PPO, DPO, and KTO in fine-grained constraints.

Copied to clipboard

Yuxiang Zhang, Jing Chen, Junjie Wang, Yaxin Liu, Cheng Yang, Chufan Shi, Xinyu Zhu, Zihao Lin, Hanwen Wan, Yujiu Yang, Tetsuya Sakai, Tian Feng, Hayato Yamana

Alleviating Sparsity of Open Knowledge Graphs with Ternary Contrastive Learning (2022.findings-emnlp)

Challenge:	Currently, tool-augmented large language models (LLMs) only achieve total scores of 45.3 and 37.0, respectively, on a scale of 100.
Approach:	They propose a multi-level diagnostic process to assess the LLM's hallucinations through two perspectives: depth and breadth.
Outcome:	The proposed diagnostic process assesses the hallucinations of large language models through two perspectives: depth and breadth.

Copied to clipboard

Qian Li, Shafiq Joty, Daling Wang, Shi Feng, Yifei Zhang

Latent Inter-User Difference Modeling for LLM Personalization (2025.emnlp-main)

Challenge:	Existing approaches to learning KG triplets ignore ternary propagation patterns and ignore zero-shot, few-shot and synonymity problems.
Approach:	They propose a framework for contrastive learning based on ternary propagation patterns among head, relation and tail.
Outcome:	Experiments on benchmarks show that TernaryCL is superior to state-of-the-art models.

Copied to clipboard

Yilun Qiu, Tianhao Shi, Xiaoyan Zhao, Fengbin Zhu, Yang Zhang, Fuli Feng

Efficient Multi-Agent System Training with Data Influence-Oriented Tree Search (2026.acl-long)

Challenge:	Large language models (LLMs) are increasingly integrated into users’ daily lives, leading to a growing demand for personalized outputs.
Approach:	They propose a framework that models inter-user differences in the latent space instead of relying on language-based prompts.
Outcome:	The proposed framework outperforms baseline methods on personalized review generation.

Copied to clipboard

Wentao Shi, Zichun Yu, Fuli Feng, Xiangnan He, Chenyan Xiong

Contrastive Learning with Generated Representations for Inductive Knowledge Graph Embedding (2023.findings-acl)

Challenge:	Large Language Model (LLM) based multi-agent systems (MAS) have high potential for tackling complex tasks through collaborative intelligence.
Approach:	They propose a framework that incorporates influence scores to guide tree search and data selection in data synthesis.
Outcome:	The proposed framework incorporates influence scores to guide tree search and data selection in data synthesis.

Copied to clipboard

Qian Li, Shafiq Joty, Daling Wang, Shi Feng, Yifei Zhang, Chengwei Qin

AnnaAgent: Dynamic Evolution Agent System with Multi-Session Memory for Realistic Seeker Simulation (2025.findings-acl)

Challenge:	Existing methods for inductive knowledge Graphs are limited by sparsity and implicit transfer.
Approach:	They propose a Contrastive Learning framework with graph guided Variational autoencoder on Meta-KGs to capture and transfer entities.
Outcome:	The proposed framework outperforms state-of-the-art methods with extensive experiments.

Copied to clipboard

Ming Wang, Peidong Wang, Lin Wu, Xiaocui Yang, Daling Wang, Shi Feng, Yuxin Chen, Bixuan Wang, Yifei Zhang

Learning to Use Tools via Cooperative and Interactive Agents (2024.findings-emnlp)

Challenge:	Existing models of seeker simulations are limited by the cost and ethical concerns of involving real seekers in mental health research.
Approach:	They propose an emotional and cognitive dynamic agent system equipped with tertiary memory to enable dynamic control of the simulator's configurations.
Outcome:	The proposed system achieves more realistic seeker simulation compared to baselines.

Copied to clipboard

Zhengliang Shi, Shen Gao, Xiuyi Chen, Yue Feng, Lingyong Yan, Haibo Shi, Dawei Yin, Pengjie Ren, Suzan Verberne, Zhaochun Ren

InsBank: Evolving Instruction Subset for Ongoing Alignment (2025.findings-emnlp)

Challenge:	Existing methods for large language models (LLMs) use one agent to iterate and execute tools, but they suffer from performance degradation when addressing practical tasks.
Approach:	They propose a tool learning framework that coordinates three specialized agents for tool selection, tool execution, and action calibration separately.
Outcome:	The proposed framework outperforms baseline models on three datasets with 14% higher success rate.

Copied to clipboard

Jiayi Shi, Yiwei Li, Shaoxiong Feng, Peiwen Yuan, Xinglin Wang, Yueqi Zhang, Chuyi Tan, Boyuan Pan, Huan Ren, Yao Hu, Kan Li

DeepMed: Building a Medical DeepResearch Agent via Multi-hop Med-Search Data and Turn-Controlled Agentic Training & Inference (2026.findings-acl)

Challenge:	Recent studies emphasize that quality and diversity of instruction data are more crucial than quantity, highlighting the need to select diverse, high-quality subsets to reduce training costs.
Approach:	They propose to use a continuously updated repository to integrate the latest valuable instruction data with a progressive evolution framework to evolve InsBank over time.
Outcome:	The proposed framework outperforms baselines in InsBank evolution and extracts budget-specific subsets.

Copied to clipboard

Zihan Wang, Hao Wang, Shi Feng, Xiaocui Yang, Daling Wang, Yiqun Zhang, Jinghao Lin, Xiaozhong Ji, Haihua Yang

EmpCRL: Controllable Empathetic Response Generation via In-Context Commonsense Reasoning and Reinforcement Learning (2024.lrec-main)

Challenge:	Medical reasoning models are constrained by parametric knowledge and can induce hallucinations and spurious attributions.
Approach:	They propose a model that uses a multi-hop med-search QA synthesis method to apply the DR paradigm in medical contexts.
Outcome:	The proposed model outperforms larger medical reasoning models on medical benchmarks.

Copied to clipboard

Mingxiu Cai, Daling Wang, Shi Feng, Yifei Zhang

Why Do More Experts Fail? A Theoretical Analysis of Model Merging (2026.acl-long)

Challenge:	Existing studies lack the perception of fine-grained dialogue emotion propagation, and have limitations in reasoning about the intentions of users on cognition, which affect the quality of empathetic response.
Approach:	They propose to use commonsense reasoning and reinforcement learning to generate empathetic response based on in-context commonsensing and contextual reasoning to broaden cognitive boundaries.
Outcome:	The proposed model outperforms state-of-the-art models in automatic and human evaluation.

Copied to clipboard

Zijing Wang, Xingle Xu, YongKang Liu, Yiqun Zhang, Peiqin Lin, Shi Feng, Daling Wang, Xiaocui Yang, Hinrich Schuetze

Challenge:	Existing methods for model merging struggle to maintain performance gains as the number of merged models increases.
Approach:	They propose a Reparameterized Heavy-Tailed method to extend the merged model’s coverage and enhance performance.
Outcome:	The proposed method extends the merged model’s coverage and enhances performance on 19 benchmarks, including knowledge-intensive and general-purpose tasks.

DNA: Denoised Neighborhood Aggregation for Fine-grained Category Discovery (2023.emnlp-main)

Copied to clipboard

Wenbin An, Feng Tian, Wenkai Shi, Yan Chen, Qinghua Zheng, QianYing Wang, Ping Chen

GenPT: Beyond Self-Report for Reliable LLM Psychometrics via Generative Projective Testing (2026.acl-long)

Challenge:	Existing methods to learn compact cluster representations from coarsely labeled data are noisy and degrade the quality of learning.
Approach:	They propose a framework that encodes semantic structures of data into the embedding space . they retrieve k-nearest neighbors of a query as positive keys to capture similarities .
Outcome:	The proposed framework can retrieve more accurate neighbors and outperform state-of-the-art models by a large margin.

Copied to clipboard

Ming Wang, Shuang Wu, Bixuan Wang, Lu Lin, Yuxin Chen, Xiaocui Yang, Daling Wang, Shi Feng, Yifei Zhang, Yufan Sun

Challenge:	Large language models (LLMs) inherit contamination from training corpora, directional bias under social-desirability framing, and limited responsiveness to context beyond the item text.
Approach:	They propose a paradigm that reformulates TAT, Rorschach, and SCT with newly generated stimuli and organises assessment as a three-stage pipeline.
Outcome:	The proposed paradigm reformulates TAT, Rorschach, and SCT with newly generated stimuli and organises assessment as a three-stage pipeline.

RAAMove: A Corpus for Analyzing Moves in Research Article Abstracts (2024.lrec-main)

Copied to clipboard

Hongzheng Li, Ruojin Wang, Ge Shi, Xing Lv, Lei Lei, Chong Feng, Fang Liu, Jinkun Lin, Yangguang Mei, Linnan Xu

PACE: Prefix-Protected and Difficulty-Aware Compression for Efficient Reasoning (2026.findings-acl)

Challenge:	RAAMove is a comprehensive multi-domain corpus dedicated to the annotation of move structures in Research Article (RA) abstracts.
Approach:	They propose a multi-domain corpus dedicated to the annotation of move structures in RA abstracts.
Outcome:	The proposed corpus is based on a human-annotated dataset and a BERT-based model to verify its effectiveness.

Copied to clipboard

Ruixiang Feng, Yuntao Wen, Silin Zhou, Ke Shi, Yifan Wang, Ran Le, Zhenwei An, Zongchao Chen, Chen Yang, Guangyue Peng, Yiming Jia, Dongsheng Wang, Tao Zhang, Lisi Chen, Yang Song, Shen Gao, Shuo Shang

Leveraging Unpaired Feedback for Long-Term LLM-based Recommendation Tuning (2025.findings-emnlp)

Challenge:	Existing LRMs often suffer from "overthinking" and excessively long reasoning traces . a dual-level framework for length compression of LRM is proposed .
Approach:	They propose a framework for prefix-protected and difficulty-aware compression under hierarchical supervision.
Outcome:	The proposed framework reduces token usage while improving accuracy on math benchmarks.

Copied to clipboard

Jizhi Zhang, Chongming Gao, Wentao Shi, Xin Chen, Jingang Wang, Xunliang Cai, Fuli Feng

Challenge:	a recent study highlights unpaired feedback as a key challenge for long-term LLM-based recommenders . unpaired user feedback is crucial for improving LLMs in dynamic user environments, authors say .
Approach:	They propose a framework that incorporates unpaired feedback into LLMs to improve long-term recommendation performance.
Outcome:	The proposed framework improves long-term recommendation performance by incorporating unpaired feedback without requiring paired supervision.

TOOL-ED: Enhancing Empathetic Response Generation with the Tool Calling Capability of LLM (2025.coling-main)

Copied to clipboard

Huiying Cao, Yiqun Zhang, Shi Feng, Xiaocui Yang, Daling Wang, Yifei Zhang

Text Generation

Challenge:	Empathetic conversation is a crucial characteristic in daily conversations between individuals.
Approach:	They propose an Emotional Knowledge Tool Calling framework which encapsulates commonsense knowledge bases as empathetic tools, enabling LLMs to integrate external knowledge flexibly.
Outcome:	The proposed framework can generate empathetic responses effectively on the TOOL-ED dataset.

Answering Narrative-Driven Recommendation Queries via a Retrieve–Rank Paradigm and the OCG-Agent (2025.emnlp-main)

Copied to clipboard

Yunxiao Shi, Haoning Shang, Xing Zi, Wujiang Xu, Yue Feng, Min Xu

Question Answering

Challenge:	Existing approaches to generate narrative-driven recommendation are based on large language models (LLMs) but the RAG paradigm is inherently ill-suited for such special queries.
Approach:	They propose a novel retrieve-rank paradigm that generatively retrieves structurally adaptive and semantically aligned candidates, ensuring both extensive candidate coverage and high-quality information.
Outcome:	The proposed paradigm outperforms the existing paradigm and the existing one under real-world scenarios.

MoLAN: A Unified Modality-Aware Noise Dynamic Editing Framework for Multimodal Sentiment Analysis (2026.findings-acl)

Copied to clipboard

Xingle Xu, YongKang Liu, Dexian Cai, Shi Feng, Xiaocui Yang, Daling Wang, Yifei Zhang

Concealed Data Poisoning Attacks on NLP Models (2021.naacl-main)

Challenge:	Existing approaches to multimodal sentiment analysis treat entire modality as an independent unit for feature enhancement or denoising, which often suppresses redundant noise at the cost of weakening critical information.
Approach:	They propose a ModaLity-aware noise dynAmic editiNg framework that performs modality-awful block partitioning by dividing features of each modality into multiple blocks.
Outcome:	Experiments on five models and four datasets show that MoLAN+ achieves the state-of-the-art performance.

Copied to clipboard

Eric Wallace, Tony Zhao, Shi Feng, Sameer Singh

SocialEval: Evaluating Social Intelligence of Large Language Models (2025.acl-long)

Challenge:	In contrast, adversarial attacks can cause model errors by modifying inputs, such as the universal triggers attack.
Approach:	They propose a data poisoning attack that allows an adversary to control model predictions whenever a desired trigger phrase is present in the input.
Outcome:	The proposed attack can cause model errors by modifying inputs, but it can also cause extra human annotation.

Copied to clipboard

Jinfeng Zhou, Yuxuan Chen, Yihan Shi, Xuanming Zhang, Leqi Lei, Yi Feng, Zexuan Xiong, Miao Yan, Xunzhi Wang, Yaru Cao, Jianing Yin, Shuai Wang, Quanyu Dai, Zhenhua Dong, Hongning Wang, Minlie Huang

PLAWBENCH: A Rubric-Based Benchmark for Evaluating LLMs in Real-World Legal Practice (2026.acl-long)

Challenge:	Existing work on LLMs does not address their social intelligence (SI) and their discrepancy with humans.
Approach:	They propose a script-based bilingual SI benchmark that integrates outcome-oriented goal achievement evaluation and process-oriented interpersonal ability evaluation by manually crafting narrative scripts.
Outcome:	The proposed model is based on a script-based bilingual evaluation paradigm that integrates outcome- and process-oriented evaluation by manually crafting narrative scripts.

Copied to clipboard

Yuzhen Shi, Huanghai Liu, Yiran HU, Song Gaojie, Xu Xinran, Yubo Ma, Tianyi Tang, Li Zhang, Qingjing Chen, Feng Di, Wenbo Lv, Weiheng Wu, Kexin Yang, Sen Yang, Wei Wang, Rongyao Shi, Qiu Yuanyang, Yuemeng Qi, Zhang Jingwen, Sui Xiaoyu, Yifan Chen, Zhang Yi, An Yang, Bowen Yu, Dayiheng Liu, Junyang Lin, Weixing Shen, Bing Zhao, Charles L. A. Clarke, HU Wei

Recurrent Knowledge Identification and Fusion for Language Model Continual Learning (2025.acl-long)

Challenge:	Existing benchmarks for large language models (LLMs) are coarse, single-dimensional metrics and do not explicitly assess fine-grained legal reasoning.
Approach:	They propose a Practical Law Benchmark to evaluate large language models in real-world legal practice scenarios.
Outcome:	The proposed model is based on 850 questions and 13 scenarios with expert-designed evaluation rubrics.

Copied to clipboard

Yujie Feng, Xujia Wang, Zexin Lu, Shenghong Fu, Guangyuan Shi, Yongxin Xu, Yasha Wang, Philip S. Yu, Xu Chu, Xiao-Ming Wu

Stealthy Jailbreak Attacks on Large Language Models via Benign Data Mirroring (2025.naacl-long)

Challenge:	Continual learning (CL) is crucial for large language models without costly retraining.
Approach:	They propose a framework for recurrent knowledge identification and fusion that enables dynamic estimation of parameter importance distributions to enhance knowledge transfer.
Outcome:	The proposed framework mitigates catastrophic forgetting and enhances knowledge transfer.

Copied to clipboard

Honglin Mu, Han He, Yuxin Zhou, Yunlong Feng, Yang Xu, Libo Qin, Xiaoming Shi, Zeming Liu, Xudong Han, Qi Shi, Qingfu Zhu, Wanxiang Che

HiFT: A Hierarchical Full Parameter Fine-Tuning Strategy (2024.emnlp-main)

Challenge:	Existing black-box jailbreak methods often rely on model feedback . existing methods may be intercepted by content moderators during the search process .
Approach:	They propose a method that guides malicious prompt construction by local training a mirror model of the target black-box model through benign data distillation.
Outcome:	The proposed method achieves a 92% attack success rate and 80% stealth rate on a subset of AdvBench.

Copied to clipboard

YongKang Liu, Yiqun Zhang, Qian Li, Tong Liu, Shi Feng, Daling Wang, Yifei Zhang, Hinrich Schuetze

Exploring and Adapting Chinese GPT to Pinyin Input Method (2022.acl-long)

Challenge:	Existing approaches to fine-tuning language models use zeroth-order optimizers to conserve GPU memory.
Approach:	They propose a full-parameter fine-tuning strategy which updates a subset of parameters at each training step.
Outcome:	The proposed approach reduces the amount of gradients and optimizer state parameters residing in GPU memory at the same time, thereby reducing GPU memory usage.

Copied to clipboard

Minghuan Tan, Yong Dai, Duyu Tang, Zhangyin Feng, Guoping Huang, Jing Jiang, Jiwei Li, Shuming Shi

SemanticCamo: Jailbreaking Large Language Models through Semantic Camouflage (2025.findings-acl)

Challenge:	a frozen GPT can generate state-of-the-art performance on perfect pinyin, but performance drops when input includes abbreviated pinyan, which links to even larger number of Chinese characters.
Approach:	They propose to use Chinese GPT to generate fluent sentences using abbreviated pinyin.
Outcome:	The proposed approach improves on abbreviated pinyin across all domains.

Copied to clipboard

Jihui Yan, Xiaocui Yang, Daling Wang, Shi Feng, Yifei Zhang, Yinzhi Zhao

MUSE: A Multimodal Conversational Recommendation Dataset with Scenario-Grounded User Profiles (2025.findings-acl)

Challenge:	Large Language Models (LLMs) have made safety issues of LLMs more prominent and critical.
Approach:	They propose a framework which attacks LLMs through semantic camouflage and replaces unsafe content with semantic features to conceal malicious intent .
Outcome:	The proposed framework outperforms existing models in over 80% of cases and is highly effective against various defenses.

Copied to clipboard

Zihan Wang, Xiaocui Yang, YongKang Liu, Shi Feng, Daling Wang, Yifei Zhang

SAD: A Large-Scale Strategic Argumentative Dialogue Dataset (2026.acl-long)

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.

Copied to clipboard

YongKang Liu, Jiayang Yu, Mingyang Wang, Yiqun Zhang, Ercong Nie, Shi Feng, Daling Wang, Kaisong Song, Hinrich Schuetze

A Co-Attention Neural Network Model for Emotion Cause Analysis with Emotional Context Awareness (D18-1)

Challenge:	Argumentation is a key part of human reasoning and decision-making . existing argumentative corpora focus on single-turn settings, but multi-turn dialogues are often realized as multi-turned dialogues .
Approach:	They present a dataset for strategic multi-turn argumentation dialogues . they annotate each utterance with five strategy types, allowing multiple strategies per utterrance .
Outcome:	The proposed dataset shows that explicit prompting improves fluency, stylistic coherence and persuasiveness.

Copied to clipboard

Xiangju Li, Kaisong Song, Shi Feng, Daling Wang, Yifei Zhang

Sentiment Analysis

Challenge:	Existing methods ignore the contexts around the emotion word which can provide an emotion cause clue.
Approach:	They propose a co-attention neural network model for emotion cause analysis with emotional context awareness.
Outcome:	The proposed model outperforms the state-of-the-art methods.

Beyond One-Size-Fits-All: Tailored Benchmarks for Efficient Evaluation (2025.acl-long)

Copied to clipboard

Peiwen Yuan, Yueqi Zhang, Shaoxiong Feng, Yiwei Li, Xinglin Wang, Jiayi Shi, Chuyi Tan, Boyuan Pan, Yao Hu, Kan Li

Boundary Detection with BERT for Span-level Emotion Cause Analysis (2021.findings-acl)

Challenge:	Existing efficient methods estimate performance of models on large benchmarks, but these methods rely on the assumption that target models have high prediction consistency with source models.
Approach:	They propose a method that conducts customized evaluation tailored to each target model.
Outcome:	The proposed method reduces the MAE of estimates by 31.4% on benchmarks across 300 models.

Copied to clipboard

Xiangju Li, Wei Gao, Shi Feng, Yifei Zhang, Daling Wang

Reverse Question Answering: Can an LLM Write a Question so Hard (or Bad) that it Can’t Answer? (2025.naacl-short)

Challenge:	Emotion cause analysis (ECA) is an emerging topic in natural language processing, which aims to identify the reasons behind a given emotion.
Approach:	They propose to detect the precise boundaries of text spans conveying accurate emotion causes from the given context by a sequence labeling and position identification problem.
Outcome:	The proposed methods outperform existing models on two benchmark datasets on the emotion cause analysis task.

Copied to clipboard

Nishant Balepur, Feng Gu, Abhilasha Ravichander, Shi Feng, Jordan Lee Boyd-Graber, Rachel Rudinger

Human-Centered Evaluation of Explanations (2022.naacl-tutorials)

Challenge:	Question answering (QA) is a popular task, but we test both separately . a recent study found that LLMs are less accurate in numerical RQA than RQA .
Approach:	We run 16 LLMs on QA and RQA with trivia questions/answers . they find question and answer types that lead to RQA errors and suggest improvements .
Outcome:	The results show that LLMs are less accurate in RQA for numerical answers than RQA . RQA errors correlate with question difficulty and inversely correlate with answer frequencies .

Copied to clipboard

Jordan Boyd-Graber, Samuel Carton, Shi Feng, Q. Vera Liao, Tania Lombrozo, Alison Smith-Renner, Chenhao Tan

Pixel-Level Reasoning Segmentation via Multi-turn Conversations (2025.acl-long)

Challenge:	This tutorial will provide an overview of human-centered evaluations of explanations .
Approach:	This tutorial will provide an overview of human-centered evaluations of explanations . it will introduce the psychological foundation of explanation and types of NLP explanations.
Outcome:	This tutorial will provide an overview of human-centered evaluations of explanations . it will cover the two categories of evaluation: evaluation based on human-annotated explanations and evaluation with human-subjects studies.

Copied to clipboard

Dexian Cai, Xiaocui Yang, YongKang Liu, Daling Wang, Shi Feng, Yifei Zhang, Soujanya Poria

Structured Attention for Unsupervised Dialogue Structure Induction (2020.emnlp-main)

Challenge:	Existing visual perception systems focus on region-level segmentation in single-turn dialogues . existing systems cannot reason at the pixel level and comprehend dynamic user intent .
Approach:	They propose a task that tracks evolving user intent via multi-turn interactions for fine-grained segmentation.
Outcome:	The proposed method outperforms existing baselines in segmentation and reasoning metrics.

Copied to clipboard

Liang Qiu, Yizhou Zhao, Weiyan Shi, Yuan Liang, Feng Shi, Tao Yuan, Zhou Yu, Song-Chun Zhu

Learning to Execute Actions or Ask Clarification Questions (2022.findings-naacl)

Challenge:	Using structured attention, a model can learn dialogue structure in unsupervised fashion.
Approach:	They propose to incorporate structured attention layers into a Variational Recurrent Neural Network model with discrete latent states to learn dialogue structure in an unsupervised fashion.
Outcome:	The proposed model learns semantic structures similar to templates used to generate a dialogue corpus on two-party datasets and on multi-party dialogues, disentangling dialogues without human annotation.

Copied to clipboard

Zhengxiang Shi, Yue Feng, Aldo Lipani

Data Swarms: Optimizable Generation of Synthetic Evaluation Data (2026.findings-acl)

Challenge:	Existing work on Minecraft Corpus Dataset only learns to execute instructions neglecting the importance of asking for clarifications.
Approach:	They propose to annotate all builder utterances into eight types, including clarification questions, and propose a builder agent model capable of determining when to ask or execute instructions.
Outcome:	The proposed model outperforms existing models on the collaborative building task with a substantial improvement.

Copied to clipboard

Shangbin Feng, Yike Wang, Weijia Shi, Yulia Tsvetkov

DialogConv: A Lightweight Fully Convolutional Network for Multi-view Response Selection (2022.emnlp-main)

Challenge:	Extensive experiments demonstrate that Data Swarms outperforms eight data generation baselines across five evaluation objectives.
Approach:	They propose an algorithm to optimize the generation of synthetic evaluation data and advance quantitative desiderata of LLM evaluation.
Outcome:	The proposed algorithm outperforms baseline evaluations and Adversarial Swarms generates harder data while learning from such data.

Copied to clipboard

Yongkang Liu, Shi Feng, Wei Gao, Daling Wang, Yifei Zhang

How Pre-trained Word Representations Capture Commonsense Physical Comparisons (D19-60)

Challenge:	Existing retrieval-based dialogue systems suffer from slow inference or huge number of parameters.
Approach:	They propose a lightweight fully convolutional architecture for response selection using convolution.
Outcome:	The proposed architecture extracts matching features of context and response from 3D views.

Copied to clipboard

Pranav Goel, Shi Feng, Jordan Boyd-Graber

DR-HM: Distill-then-Reinforce Training with Cognition-Aware Data Synthesis for Harmful Meme Detection (2026.findings-acl)

Challenge:	Pre-trained word representations capture common sense on physical properties such as size and weight.
Approach:	They investigate whether pre-trained representations capture comparisons and find they have higher accuracy than previous approaches.
Outcome:	The proposed models learn a consistent ordering over all the objects in the comparisons.

Copied to clipboard

Zihan Cheng, Jianxiang Ma, Xiaocui Yang, Peidong Wang, Wen Zhang, Shi Feng, Daling Wang, Yifei Zhang, Mingfu Zhang

R3: End-to-End Reasoning-based Planning for Multi-step Retrosynthesis via Reinforcement Learning (2026.acl-long)

Challenge:	Current methods for harmful meme detection lack the knowledge required to identify such hate . current methods lack the ability to identify cultural stereotypes and visual metaphors .
Approach:	They propose a framework that decomposes meme analysis into a human-inspired reasoning process . they propose DR-HM to transfer knowledge from closed-source models while mitigating biases .
Outcome:	The proposed framework outperforms existing methods on three benchmark datasets.

Copied to clipboard

YiFei Wang, Qizhi Pei, Jiangtao Feng, Yuntian Shi, Yi Duan, Lihao Wang, Lei Bai, Lijun Wu, Wei-Ying Ma, Hao Zhou

Whose Boat Does it Float? Improving Personalization in Preference Tuning via Inferred User Personas (2025.acl-long)

Challenge:	Experimental results show that R3 is a superior alternative to traditional search algorithms for multistep retrosynthesis planning.
Approach:	They propose a framework that reformulates multistep retrosynthetic planning as a generative reasoning task.
Outcome:	The proposed framework achieves state-of-the-art Top-1 accuracy of 43.7% on retrobench . it leverages Large Language Models to reformulate multistep retrosynthesis as a generative reasoning task.

Copied to clipboard

Nishant Balepur, Vishakh Padmakumar, Fumeng Yang, Shi Feng, Rachel Rudinger, Jordan Lee Boyd-Graber

Nature-Inspired Population-Based Evolution of Large Language Models (2026.acl-long)

Challenge:	a recent study shows that LLMs can't tailor outputs to users with uncommon preferences . despite the success of persona inference, we may need debiasing and abstention.
Approach:	They propose to use preference data to infer needs and interests of users who prefer either output . they argue that training on preference data augmented with PI boosts personalization .
Outcome:	The proposed method can be used to improve personalization with less privacy concerns.

Copied to clipboard

Yiqun Zhang, Peng Ye, Xiaocui Yang, Shi Feng, Shufei Zhang, Lei Bai, Wanli Ouyang, Shuyue Hu

TUNA: Comprehensive Fine-grained Temporal Understanding Evaluation on Dense Dynamic Videos (2025.acl-long)

Challenge:	a new framework for population-based evolution of large language models is emerging . a population-driven evolution of LLMs is a key component of evolution, authors say .
Approach:	They propose a framework that allows for population-based evolution of large language models . they start with a population of parent LLMs and allow this population to evolve .
Outcome:	The proposed framework outperforms existing methods on 12 datasets.

Copied to clipboard

Fanheng Kong, Jingyuan Zhang, Hongzhi Zhang, Shi Feng, Daling Wang, Linhao Yu, Xingguang Ji, Yu Tian, V. W., Fuzheng Zhang

Visual Data in NLP

Challenge:	Existing benchmarks for video understanding often focus on specific aspects, overlooking the holistic nature of video content.
Approach:	They propose a temporal-oriented benchmark for fine-grained understanding on dense dynamic videos with two complementary tasks: captioning and QA.
Outcome:	The proposed model performs well on diverse video scenarios and dynamic videos, with interpretable and robust evaluation criteria.

Misleading Failures of Partial-input Baselines (P19-1)

Copied to clipboard

Shi Feng, Eric Wallace, Jordan Boyd-Graber

Challenge:	Recent work establishes dataset difficulty and removes annotation artifacts via partial-input baselines.
Approach:	They propose to use partial-input baselines to establish dataset difficulty . they show how trivial patterns only visible in the full input can evade partial-output baseline .
Outcome:	The proposed model can solve 15% of previously-thought "hard" examples.

Decoding in Latent Spaces for Efficient Inference in LLM-based Recommendation (2025.findings-emnlp)

Copied to clipboard

Chengbing Wang, Yang Zhang, Zhicheng Wang, Tianhao Shi, Keqin Bao, Fuli Feng, Tat-Seng Chua

SciCustom: A Framework for Custom Evaluation of Scientific Capabilities in Large Language Models (2026.acl-long)

Challenge:	Light Latent-space Decoding (L2D) is an efficient and efficient latent- space decoding method.
Approach:	They propose to bypass language-space decoding by matching candidate items with LLM's internal thought representations in the latent space.
Outcome:	The proposed method is 10x faster than language-space decoding while maintaining or enhancing performance.

Copied to clipboard

Yiyang Gu, Junwei Yang, Junyu Luo, Ye Yuan, Bin Feng, Yingce Xia, Shufang Xie, Kaili Liu, Bohan Wu, Qi Shi, Haoran Li, Beier Xiao, Zhiping Xiao, Xiao Luo, Weizhi Zhang, Philip S. Yu, Zequn Liu, Ming Zhang

Learning to Improve Persona Consistency in Multi-party Dialogue Generation via Text Knowledge Enhancement (2022.coling-1)

Challenge:	Existing evaluations of large language models fail to reflect fine-grained capabilities . existing benchmarks are manually curated or domain-generic, limiting scalability and alignment with real use cases.
Approach:	They propose a framework that allows custom construction of benchmarks from large-scale scientific data to evaluate application-specific scientific capabilities in LLMs.
Outcome:	The proposed framework reveals fine-grained differences in scientific capabilities that standard benchmarks overlook . it allows custom construction of benchmarks from large-scale scientific data to evaluate application-specific capabilities in LLMs.

Copied to clipboard

Dongshi Ju, Shi Feng, Pengcheng Lv, Daling Wang, Yifei Zhang

KC-ISA: An Implicit Sentiment Analysis Model Combining Knowledge Enhancement and Context Features (2022.coling-1)

Challenge:	Existing methods suffer from incomprehensive persona tags that have unique and obscure meanings to describe human’s personality.
Approach:	They propose a graph convolution network model with addressee selecting mechanism that integrates personas, dialogue utterances, and external text knowledge in a unified graph.
Outcome:	The proposed model outperforms baselines by large margins and improves persona consistency in the generated responses.

Copied to clipboard

Minghao Xu, Daling Wang, Shi Feng, Zhenfei Yang, Yifei Zhang

Sentiment Analysis

Challenge:	Existing research results on explicit sentiment analysis are limited . implicit sentiment analysis is a process of analyzing text based on whether it contains explicit sentiment words.
Approach:	They propose a model that integrates external knowledge and contextual features . they use a knowledge graph to supplement implicit sentiment expression .
Outcome:	The proposed model can achieve better results on the SMP2019 implicit sentiment analysis dataset.

Measuring Inductive Biases of In-Context Learning with Underspecified Demonstrations (2023.acl-long)

Copied to clipboard

Chenglei Si, Dan Friedman, Nitish Joshi, Shi Feng, Danqi Chen, He He

Improved Visual Story Generation with Adaptive Context Modeling (2023.findings-acl)

Challenge:	In-context learning is an important paradigm for adapting large language models to new tasks . but the generalization behavior of ICL remains poorly understood .
Approach:	They characterize the feature biases of large language models by constructing underspecified demonstrations . they find that LLMs exhibit clear feature bias, and they evaluate interventions .
Outcome:	The proposed model prefers the "default" task features over distractor features more often than the base model.

Copied to clipboard

Zhangyin Feng, Yuchen Ren, Xinmiao Yu, Xiaocheng Feng, Duyu Tang, Shuming Shi, Bing Qin

JX4MEI: Multimodal Semantically-Enhanced LLM for Joint Multimodal Emotion-Intent Explanation and Classification (2026.findings-acl)

Challenge:	a recent work shows that diffusion models generate images of high resolution and semantic consistency to text prompts.
Approach:	They propose a method that uses adaptive context modeling to improve leading system . they evaluate their method on pororoSV and FlintstonesSV datasets .
Outcome:	The proposed method achieves state-of-the-art FID scores on pororo and Flintstones datasets.

Copied to clipboard

YiJie Huang, Xiaocui Yang, Shi Feng, Daling Wang, Yifei Zhang, Ning Yuan, Zhuoyue Jia, Wen Zhang

Information Retrieval

Challenge:	Existing multimodal emotion and intent recognition tasks focus on classification, not rationale and intrinsic connections between these states.
Approach:	They propose a task that requires models to jointly predict emotion and intent while generating natural language explanations for why they co-occur.
Outcome:	The proposed model outperforms baseline models in prediction and explanation generation.

AJ-Bench: Benchmarking Agent-as-a-Judge for Environment-Aware Evaluation (2026.findings-acl)

Copied to clipboard

Wentao Shi, Yu Wang, Yuyang Zhao, Yuxin Chen, Fuli Feng, Xueyuan Hao, Xi Su, Qi GU, Hui Su, Xunliang Cai, Xiangnan He

Challenge:	Existing approaches to verify agent behaviors in complex environments rely on rule-based verifiers or LLM-as-a-Judge models.
Approach:	They propose a benchmark to evaluate Agent-as-a-Judge across three domains . the benchmark covers search, data systems, and graphical user interfaces - with 155 tasks and 516 trajectories .
Outcome:	The proposed benchmark outperforms existing benchmarks in search, data systems, and GUI domains while revealing open challenges in agent-based verification.

FlowSearch: Advancing Deep Research with Dynamic Structured Knowledge Flow (2026.acl-long)

Copied to clipboard

Yusong Hu, Runmin Ma, Yue Fan, Jinxin Shi, Zongsheng Cao, Yuhao Zhou, Jiakang Yuan, Shuaiyu Zhang, Shiyang Feng, Xiangchao Yan, Shufei Zhang, Wenlong Zhang, Lei Bai, Bo Zhang

Multimodal Sentiment Detection Based on Multi-channel Graph Neural Networks (2021.acl-long)

Challenge:	FlowSearch is a multi-agent framework that actively constructs and evolves a dynamic structured knowledge flow to drive subtask execution and reasoning.
Approach:	They propose a multi-agent framework that actively constructs and evolves a dynamic structured knowledge flow to drive subtask execution and reasoning.
Outcome:	The proposed framework achieves competitive performance on GAIA, HLE, GPQA and TRQA benchmarks and is available to download.

Copied to clipboard

Xiaocui Yang, Shi Feng, Yifei Zhang, Daling Wang

ResLoRA: Identity Residual Mapping in Low-Rank Adaption (2024.findings-acl)

Challenge:	Existing studies only considered the representation of a single image-text post . Fig. 1 shows that multimodal sentiment expressions have global characteristics .
Approach:	They propose a multi-channel Graph Neural Networks with Sentiment-awareness approach for image-text sentiment detection.
Outcome:	The proposed approach is effective for image-text sentiment detection on three publicly available datasets.

Copied to clipboard

Shuhua Shi, Shaohan Huang, Minghui Song, Zhoujun Li, Zihan Zhang, Haizhen Huang, Furu Wei, Weiwei Deng, Feng Sun, Qi Zhang

Challenge:	Low-rank adaptation (LoRA) is one of the most popular parameter-efficient fine-tuning methods.
Approach:	They propose a low-rank adaptation method that adds residual paths during training and merges them together during inference to achieve better results.
Outcome:	The proposed method achieves 2.5x faster convergence speed and improves performance by 14.3% on NLG, NLU, and text-to-image tasks.

SSMLoRA: Enhancing Low-Rank Adaptation with State Space Model (2025.naacl-long)

Copied to clipboard

Jiayang Yu, Yihang Zhang, Bin Wang, Peiqin Lin, YongKang Liu, Shi Feng

Large Language Models Help Humans Verify Truthfulness – Except When They Are Convincingly Wrong (2024.naacl-long)

Challenge:	Fine-tuning requires substantial computational resources and is prone to overfitting when applied to small datasets.
Approach:	They propose a parameter-efficient fine-tuning method that integrates a State Space Model (SSM) to interconnect low-rank matrices.
Outcome:	The proposed method achieves comparable performance to LoRA on the general language understanding evaluation (GLUE) benchmark while using only half the parameters.

Copied to clipboard

Chenglei Si, Navita Goyal, Tongshuang Wu, Chen Zhao, Shi Feng, Hal Daumé Iii, Jordan Boyd-Graber

ProxyQA: An Alternative Framework for Evaluating Long-Form Text Generation with Large Language Models (2024.acl-long)

Challenge:	Large Language Models (LLMs) are increasingly used for accessing information on the web.
Approach:	They conduct experiments with 80 crowdworkers to compare LLMs with search engines . they ask LLM to provide contrastive information to reduce over-reliance on LLM .
Outcome:	The results show that LLMs can outperform search engines but not LLM explanations . the study shows that LMS explanations are not reliable replacements for reading retrieved passages compared to search engines alone.

Copied to clipboard

Haochen Tan, Zhijiang Guo, Zhan Shi, Lu Xu, Zhili Liu, Yunlong Feng, Xiaoguang Li, Yasheng Wang, Lifeng Shang, Qun Liu, Linqi Song

Active Example Selection for In-Context Learning (2022.emnlp-main)

Challenge:	Existing evaluation methods for large language models are labor-intensive and lack efficiency.
Approach:	They propose a framework dedicated to assessing long-text generation that includes in-depth human-curated meta-questions spanning various domains . they use a set of proxy-quests with pre-annotated answers to assess the content's quality by incorporating the generated texts as contextual background.
Outcome:	The proposed framework assesses the quality of long-text content by matching it with references through human evaluation or automated metrics.

Copied to clipboard

Yiming Zhang, Shi Feng, Chenhao Tan

Global-Local Modeling with Prompt-Based Knowledge Enhancement for Emotion Inference in Conversation (2023.findings-eacl)

Challenge:	In-context learning performance is unstable across samples of examples, suggesting the idiosyncrasies of how language models acquire information.
Approach:	They propose a reinforcement learning algorithm for identifying generalizable policies to select demonstration examples and propose 'in-context learning' performance can be highly unstable across samples of examples, suggesting the idiosyncrasies of how language models acquire information.
Outcome:	The proposed model can perform tasks with examples with a 5.8% improvement on GPT-2 and GPT-3, but the improvement diminishes on larger models, suggesting emerging capabilities of large language models.

Copied to clipboard

Renxi Wang, Shi Feng

Speculative Decoding for Multi-Sample Inference (2025.findings-emnlp)

Challenge:	Existing studies on emotion recognition focus on recognizing emotions through a speaker’s utterance, while research on emotion inference predicts emotions of addressees through previous utterations.
Approach:	They propose a global-local modeling method based on recurrent neural networks and pre-trained language models to do emotion inference in conversation.
Outcome:	The proposed method achieves state-of-the-art on three datasets.

Copied to clipboard

Yiwei Li, Jiayi Shi, Shaoxiong Feng, Peiwen Yuan, Xinglin Wang, Yueqi Zhang, Ji Zhang, Chuyi Tan, Boyuan Pan, Yao Hu, Kan Li

Improving Role-Oriented Dialogue Summarization with Interaction-Aware Contrastive Learning (2024.lrec-main)

Challenge:	Speculative decoding method exploits consensus of parallel reasoning paths to synthesize high-quality draft tokens without auxiliary models or external databases.
Approach:	They propose a speculative decoding method that exploits the consensus of parallel reasoning paths to synthesize high-quality draft tokens without auxiliary models or external databases.
Outcome:	The proposed method exploits the intrinsic consensus of parallel reasoning paths to synthesize high-quality draft tokens without auxiliary models or databases.

Copied to clipboard

Weihong Guan, Shi Feng, Daling Wang, Faliang Huang, Yifei Zhang, Yuan Cui

Self-Improvement Towards Pareto Optimality: Mitigating Preference Conflicts in Multi-Objective Alignment (2025.findings-acl)

Challenge:	Existing methods for encoding dialogues do not capture interaction information between roles, thus ignore interaction-related key information.
Approach:	They propose a contrastive learning based interaction-aware model for the role-oriented dialogue summarization namely CIAM and use it to train the decoder to learn role-level interaction.
Outcome:	The proposed model captures interaction information between different roles and produces informative summaries on two public datasets.

Copied to clipboard

Moxin Li, Yuantao Zhang, Wenjie Wang, Wentao Shi, Zhuo Liu, Fuli Feng, Tat-Seng Chua

KARL: Knowledge-Aware Retrieval and Representations aid Retention and Learning in Students (2024.emnlp-main)

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.

Copied to clipboard

Matthew Shu, Nishant Balepur, Shi Feng, Jordan Boyd-Graber

When One LLM Drools, Multi-LLM Collaboration Rules (2026.acl-long)

Challenge:	Existing student models use study data like student's past responses to predict the probability a student can recall a flashcard.
Approach:	They propose to use student models to predict recall of flashcards to build a content-aware student model that uses deep knowledge tracing, retrieval, and BERT to predict student recall.
Outcome:	The proposed content-aware student model outperforms existing student models in AUC and calibration error and is more efficient than SOTA.

Copied to clipboard

Shangbin Feng, Wenxuan Ding, Alisa Liu, Zifeng Wang, Weijia Shi, Yike Wang, Shannon Zejiang Shen, Xiaochuang Han, Hunter Lang, Chen-Yu Lee, Tomas Pfister, Yejin Choi, Yulia Tsvetkov

Challenge:	a single general-purpose LLM is not enough to produce a reliable output, argues this paper . a multi-LLM collaboration approach addresses reliability, democratization, and pluralism .
Approach:	They argue that a single general-purpose LLM is not enough to produce a reliable output . they organize existing multi-LLM collaboration methods into a hierarchy based on access and information exchange .
Outcome:	The proposed method addresses reliability, democratization, and pluralism challenges a single LLM fails to produce a reliable output.

Boosting Event Extraction with Denoised Structure-to-Text Augmentation (2023.findings-acl)

Copied to clipboard

Bo Wang, Heyan Huang, Xiaochi Wei, Ge Shi, Xiao Liu, Chong Feng, Tong Zhou, Shuaiqiang Wang, Dawei Yin

Information Extraction & Text Mining

Challenge:	Existing methods for event extraction neglect grammatical incorrectness, structure misalignment, and semantic drifting . et al., 2004; Ahn, 2006) show that the proposed method generates more diverse text representations for event extracting compared with the state-of-the-art.
Approach:	They propose a framework for event extraction that generates additional training data and iteratively selects the effective subset from the generated training data.
Outcome:	The proposed method generates more diverse representations of training data and achieves comparable results with the state-of-the-art.

DPEPO: Diverse Parallel Exploration Policy Optimization for LLM-based Agents (2026.acl-long)

Copied to clipboard

JunShuo Zhang, Chengrui Huang, Feng Guo, Zihan Li, Ke Shi, Menghua Jiang, Jiguo Yu, Shuo Shang, Shen Gao

Generalized Category Discovery with Large Language Models in the Loop (2024.findings-acl)

Challenge:	Existing approaches to large language model (LLM) agents that follow the sequential "reason-then-act" paradigm suffer from limited exploration and incomplete environmental understanding as they interact with only a single environment per step.
Approach:	They propose a paradigm that enables an agent to interact with multiple environments simultaneously and share cross-trajectory experiences.
Outcome:	The proposed paradigm achieves state-of-the-art (SOTA) success rates while maintaining comparable efficiency to strong sequential baselines.

Copied to clipboard

Wenbin An, Wenkai Shi, Feng Tian, Haonan Lin, QianYing Wang, Yaqiang Wu, Mingxiang Cai, Luyan Wang, Yan Chen, Haiping Zhu, Ping Chen

Look Within or Beyond? A Theoretical Comparison Between Parameter-Efficient and Full Fine-Tuning (2026.acl-long)

Challenge:	Generalized Category Discovery (GCD) is a crucial task that aims to recognize both known and novel categories from a set of unlabeled data.
Approach:	They propose a framework that introduces Large Language Models into the training loop to generate category names without human effort.
Outcome:	The proposed framework outperforms SOTA models on three benchmark datasets and generates accurate category names for the discovered clusters.

Copied to clipboard

YongKang Liu, Xingle Xu, Ercong Nie, Zijing Wang, Shi Feng, Daling Wang, Qian Li, Hinrich Schuetze

PlaM: Training-Free Plateau-Guided Model Merging for Better Visual Grounding in MLLMs (2026.findings-acl)

Challenge:	Parameter-Efficient Fine-Tuning (PEFT) is an alternative to Full-Parameter Fine-tuning, but its effectiveness on complex tasks such as reasoning and instruction-following remains unclear.
Approach:	They propose to use PEFT to reduce the number of trainable parameters while freezing the weights of LLMs.
Outcome:	The proposed methods perform well on standard tasks, but weaknesses on complex and adversarial settings call for new directions beyond current paradigms.

Copied to clipboard

Zijing Wang, YongKang Liu, Mingyang Wang, Ercong Nie, Deyuan Chen, Zhengjie Zhao, Shi Feng, Daling Wang, Xiaocui Yang, Yifei Zhang, Hinrich Schuetze

Teaching LLMs to Abstain across Languages via Multilingual Feedback (2024.emnlp-main)

Challenge:	Multimodal instruction fine-tuning degrades textual reasoning capability, undermining multimodal performance.
Approach:	They propose a plateau-guided model merging method that selectively injects base language model parameters into MLLMs to mitigate this degradation.
Outcome:	The proposed framework reduces multimodal instruction fine-tuning degradation by incorporating a plateau-guided model merging method into MLLMs.

Copied to clipboard

Shangbin Feng, Weijia Shi, Yike Wang, Wenxuan Ding, Orevaoghene Ahia, Shuyue Stella Li, Vidhisha Balachandran, Sunayana Sitaram, Yulia Tsvetkov

Multilinguality

Challenge:	Existing studies on LLM abstention focus on English, but they show that it can reduce the accuracy of the model by 20.5% .
Approach:	They propose to teach LLMs to abstain in the face of knowledge gaps by generating multiple feedback items in related languages.
Outcome:	Extensive experiments show that the proposed approach outperforms baselines and achieves 9.2% improvement for low-resource languages.

Answer-guided and Semantic Coherent Question Generation in Open-domain Conversation (D19-1)

Copied to clipboard

Weichao Wang, Shi Feng, Daling Wang, Yifei Zhang

CIRAG: Construction–Integration Retrieval and Adaptive Generation for Multi-hop Question Answering (2026.acl-long)

Challenge:	Existing methods for question generation suffer from dullness and deviation problem, which can lead to deviated or dull questions.
Approach:	They propose two methods to enhance semantic coherence between question and answer by using a coherent score and adversarial training to explicitly control question generation.
Outcome:	The proposed methods outperform state-of-the-art baseline algorithms with large margins in raising semantic coherent questions.

Copied to clipboard

Zili Wei, Yilin Wang, Xiaocui Yang, Shi Feng, Weidong Bao, Daling Wang, Zihan Wang, Yifei Zhang

Information Retrieval

Challenge:	Existing methods for iterative retrieval-augmented generation (iRAG) suffer from greedy single-path expansion and granularity–demand mismatch .
Approach:	They propose a model that constructs candidate triples and history-conditionally integrates them to distill core triples to generate the next-hop query.
Outcome:	The proposed model mitigates the greedy single-path expansion and granularity–demand mismatch by preserving multiple plausible evidence chains.

A Diffusion Weighted Graph Framework for New Intent Discovery (2023.emnlp-main)

Copied to clipboard

Wenkai Shi, Wenbin An, Feng Tian, Qinghua Zheng, QianYing Wang, Ping Chen

TaSL: Continual Dialog State Tracking via Task Skill Localization and Consolidation (2024.acl-long)

Challenge:	Existing methods to learn from unlabeled data generate noisy supervisory signals . current methods only rely on semantic similarities to generate supervisory signal .
Approach:	They propose a weighted DWGF framework to capture semantic similarities and structure relationships in data.
Outcome:	The proposed method outperforms state-of-the-art models on evaluation metrics across multiple benchmark datasets.

Copied to clipboard

Yujie Feng, Xu Chu, Yongxin Xu, Guangyuan Shi, Bo Liu, Xiao-Ming Wu

Advancing Sequential Numerical Prediction in Autoregressive Models (2025.acl-short)

Challenge:	Current methods for Continual Dialogue State Tracking (DST) struggle with catastrophic forgetting and knowledge transfer between tasks.
Approach:	They propose a framework for task skill localization and consolidation that enables effective knowledge transfer without relying on memory replay.
Outcome:	The proposed framework shows a 7.6% increase in Avg. JGA and 11% rise in BWT metrics over existing state-of-the-art methods.

Copied to clipboard

Xiang Fei, Jinghui Lu, Qi Sun, Hao Feng, Yanjie Wang, Wei Shi, An-Lan Wang, Jingqun Tang, Can Huang

Challenge:	Autoregressive models are the de facto choice for sequence generation tasks, but standard approaches treat digits as independent tokens and apply cross-entropy loss, overlooking the coherent structure of numerical sequences.
Approach:	They propose a novel approach to entropy loss by extending the Earth Mover’s Distance to preserve ordinal relationships between numerical values and sequence-level to penalize the overall discrepancy between predicted and actual sequences.
Outcome:	Extensive experiments show that NTIL improves numerical prediction and integrates effectively with LLMs/MLLMs.

A SMART Mnemonic Sounds like “Glue Tonic”: Mixing LLMs with Student Feedback to Make Mnemonic Learning Stick (2024.emnlp-main)

Copied to clipboard

Nishant Balepur, Matthew Shu, Alexander Hoyle, Alison Robey, Shi Feng, Seraphina Goldfarb-Tarrant, Jordan Boyd-Graber

Challenge:	a new study shows that mnemonics are not effective at matching student learning to a standardized learning model.
Approach:	They build a keyword mnemonic generator that finds mnemonics students favor in a flashcard app . they use expressed and observed preferences to find out what students think is helpful .
Outcome:	The proposed mnemonics outperform existing models in keyword mnemonics . the human writer outperformed both models in terms of keyword simplicity and explanation quality .

From Sub-Ability Diagnosis to Human-Aligned Generation: Bridging the Gap for Text Length Control via MarkerGen (2025.acl-long)

Copied to clipboard

Peiwen Yuan, Chuyi Tan, Shaoxiong Feng, Yiwei Li, Xinglin Wang, Yueqi Zhang, Jiayi Shi, Boyuan Pan, Yao Hu, Kan Li

Challenge:	Existing methods to control text length are lacking in LCTG, posing a major limitation for practical applications.
Approach:	They propose a plug-and-play approach that decomposes LCTG sub-abilities with human patterns as reference and performs detailed error analysis.
Outcome:	The proposed method significantly improves LCTG across various settings, exhibiting outstanding effectiveness and generalizability.

Language Models as Continuous Self-Evolving Data Engineers (2025.emnlp-main)

Copied to clipboard

Peidong Wang, Ming Wang, Zhiming Ma, Xiaocui Yang, Shi Feng, Daling Wang, Yifei Zhang, Kaisong Song

Mis-prompt: Benchmarking Large Language Models for Proactive Error Handling (2025.acl-long)

Challenge:	Large Language Models (LLMs) have demonstrated remarkable capabilities, yet their further evolution is often hampered by the scarcity of high-quality training data and the heavy reliance of traditional methods on expert-labeled data.
Approach:	They propose a paradigm that enables LLMs to train themselves by generating, cleaning, reviewing and annotating data with preference information.
Outcome:	The proposed model can generate, clean, review, and annotate data with preference information significantly reducing time and cost of post-training data construction.

Copied to clipboard

Jiayi Zeng, Yizhe Feng, Mengliang He, Wenhui Lei, Wei Zhang, Zeming Liu, Xiaoming Shi, Aimin Zhou

NEAT: Neuron-Based Early Exit for Large Reasoning Models (2026.findings-acl)

Challenge:	Current error-handling works are performed in a passive manner, with explicit error- handling instructions.
Approach:	They propose a new benchmark to analyze LLMs' performance on a mis-prompt benchmark and a dataset to promote further research.
Outcome:	The proposed benchmark shows that current LLMs show poor performance on proactive error handling, and that SFT improves on error handling instances.

Copied to clipboard

Kang Liu, YongKang Liu, Xiaocui Yang, Peidong Wang, Wen Zhang, Shi Feng, Yifei Zhang, Daling Wang

Revisiting Self-Consistency from Dynamic Distributional Alignment Perspective on Answer Aggregation (2025.findings-acl)

Challenge:	Existing approaches to reduce overthinking require additional rollout computation or externally labeled datasets.
Approach:	They propose a Neuron-based Early reAsoning exiT framework that monitors neuron-level activation dynamics to enable training-free early exits.
Outcome:	The proposed framework reduces the amount of reasoning steps generated by LRMs while maintaining accuracy.

Copied to clipboard

Yiwei Li, Ji Zhang, Shaoxiong Feng, Peiwen Yuan, Xinglin Wang, Jiayi Shi, Yueqi Zhang, Chuyi Tan, Boyuan Pan, Yao Hu, Kan Li

A Good Plan is Hard to Find: Aligning Models with Preferences is Misaligned with What Helps Users (2025.emnlp-main)

Challenge:	Existing studies on self-consistency show that it improves reasoning abilities by aggregating diverse stochastic samples.
Approach:	They propose a confidence-driven mechanism that dynamically calibrates temperature to align with high probability modes.
Outcome:	The proposed method outperforms fixed-diversity baselines on reasoning tasks and improves both average and best-case performance.

Copied to clipboard

Nishant Balepur, Matthew Shu, Yoo Yeon Sung, Seraphina Goldfarb-Tarrant, Shi Feng, Fumeng Yang, Rachel Rudinger, Jordan Lee Boyd-Graber

Challenge:	We test alignment methods to ensure LLMs are helpful, but they train or evaluate on what users prefer .
Approach:	They test alignment methods to ensure LLMs generate plans that help users . they get 4388 plan executions and 5584 comparisons to measure user preferences .
Outcome:	The proposed approach can be applied to the problem of user preferences and helpfulness.

Multi-Modal Multi-Granularity Tokenizer for Chu Bamboo Slips (2025.coling-main)

Copied to clipboard

Yingfa Chen, Chenlong Hu, Cong Feng, Chenyang Song, Shi Yu, Xu Han, Zhiyuan Liu, Maosong Sun

Syntactic Text Processing

Challenge:	Using a multi-modal multi-granularity tokenizer, we analyze ancient Chinese scripts . a large proportion of the characters in ancient Chinese are rare or undeciphered .
Approach:	They propose a multi-modal multi-granularity tokenizer specifically designed for ancient Chinese scripts.
Outcome:	The proposed tokenizer improves on the part-of-speech tagging task on the Chu bamboo slip script.

Learning to Explain Selectively: A Case Study on Question Answering (2022.emnlp-main)

Copied to clipboard

Shi Feng, Jordan Boyd-Graber

Explainability & Interpretability in NLP

Challenge:	Recent advances in machine learning (ML) have obstructed the use of NNs.
Approach:	They propose to learn to explain"selectively" for each decision that the user makes . they use a model to choose the best explanation from a set of candidates and update this model with feedback .
Outcome:	The proposed model improves human performance on a question-based task for experts and crowdworkers.

PVGRU: Generating Diverse and Relevant Dialogue Responses via Pseudo-Variational Mechanism (2023.acl-long)

Copied to clipboard

Yongkang Liu, Shi Feng, Daling Wang, Yifei Zhang, Hinrich Schütze

Genre Separation Network with Adversarial Training for Cross-genre Relation Extraction (D18-1)

Challenge:	Existing generative models for dialogue use the last hidden state to summarize the history of the dialogue.
Approach:	They propose a Pseudo-Variational Gated Recurrent Unit (PVGRU) that summarises the accumulated distribution variations of subsequences and builds a model based on it.
Outcome:	The proposed model can improve diversity and relevance of responses on two benchmark datasets.

Copied to clipboard

Ge Shi, Chong Feng, Lifu Huang, Boliang Zhang, Heng Ji, Lejian Liao, Heyan Huang

MTRouter: Cost-Aware Multi-Turn LLM Routing with History–Model Joint Embeddings (2026.acl-long)

Challenge:	Existing methods to extract genre-specific and genre-agnostic features require great human effort.
Approach:	They propose to use two encoders to explicitly extract genre-specific and genre-agnostic features.
Outcome:	The proposed approach outperforms the state-of-the-art by 1.7% on three distinct genres.

Copied to clipboard

Yiqun Zhang, Hao Li, Zihan Wang, Shi Feng, Xiaocui Yang, Daling Wang, Bo Zhang, Lei Bai, Shuyue Hu

SAFE-QAQ: End-to-End Slow-Thinking Audio-Text Fraud Detection via Reinforcement Learning (2026.acl-long)

Challenge:	Multi-turn, long-horizon tasks require dozens of sequential model calls per episode.
Approach:	They propose a cost-aware multi-turn LLM routing tool which encodes interaction history and candidate models into joint history–model embeddings and learns an outcome estimator from logged trajectories to predict turn-level model utility.
Outcome:	The proposed model reduces cost and performance by 58.7% on ScienceWorld and on Humanity’s Last Exam (HLE) and even reduces costs for held-out tasks.

Copied to clipboard

Peidong Wang, Zhiming Ma, Xin Dai, YongKang Liu, Shi Feng, Xiaocui Yang, Wenxing Hu, Zhihao Wang, Mingjun Pan, Li Yuan, Daling Wang

Cat-MoD: Accelerating Multimodal Alignment via Caption Token Guided Asymmetric Mixture-of-Depths (2026.acl-long)

Challenge:	Existing methods for fraud detection rely on transcribed text, lacking acoustic cues . a proposed framework for audio-based slow-thinking fraud detection eliminates transcription errors .
Approach:	They propose a framework for audio-based slow-thinking fraud detection that eliminates transcription errors and rewards slow-thought reasoning by capturing fine-grained audio details.
Outcome:	The proposed method improves accuracy, inference efficiency, and real-time processing capabilities.

Copied to clipboard

YiJie Huang, Xiaocui Yang, Shi Feng, Wen Zhang, Kaisong Song, Yifei Zhang, Daling Wang

Pathologies of Neural Models Make Interpretations Difficult (D18-1)

Challenge:	Existing query-based alignment modules enforce uniform cross-attention across all layers, leading to computational redundancy.
Approach:	They propose a framework that allows for asynchronous query-based alignment with large-scale visual features.
Outcome:	The proposed framework matches or surpasses baseline performance while reducing alignment FLOPs by approximately 37% during training and inference.

Copied to clipboard

Shi Feng, Eric Wallace, Alvin Grissom II, Mohit Iyyer, Pedro Rodriguez, Jordan Boyd-Graber

CARE-STaR: Constraint-aware Self-taught Reasoner (2025.findings-acl)

Challenge:	Existing methods for NLP use input reduction to determine a word's importance . human accuracy degrades when shown the reduced examples instead of the original .
Approach:	They propose a process that iteratively removes the least important word from an input . they show human models make the same predictions with high confidence .
Outcome:	The proposed methods expose pathological behaviors of neural models . human experiments show that reduced examples lack information to support the prediction of any label .

Copied to clipboard

Zhiliang Li, Bo Tang, Yijun Niu, Beihong Jin, Qiwen Shi, Yuchen Feng, Zhiyu Li, Jie Hu, Mingchuan Yang, Feiyu Xiong

Zero-shot Cross-domain Dialogue State Tracking via Context-aware Auto-prompting and Instruction-following Contrastive Decoding (2024.emnlp-main)

Challenge:	Recent research on instruction following has demonstrated that LLMs can handle complex instructions.
Approach:	They propose to assign constraints to different levels of constraints in instructions . they use chain-of-thought and self-taught reasoner methods to identify constraints .
Outcome:	The proposed method outperforms supervised fine-tuning (SFT) on three instruction-following benchmarks.

Copied to clipboard

Xiaoyu Dong, Yujie Feng, Zexin Lu, Guangyuan Shi, Xiao-Ming Wu

Bi-Tuning with Collaborative Information for Controllable LLM-based Sequential Recommendation (2025.acl-long)

Challenge:	Previous studies have implemented slot-based input improvements, such as schema-driven descriptions and question-answering formats, but still suffer from negative transfer for seen slots and inefficient transfer for unseen slots due to the significant source-target domain gap.
Approach:	They propose a framework that generates dynamic, context-aware slot queries to improve model transferability by penalizing deviations from the provided instructions.
Outcome:	Experiments on two datasets show that the proposed model performs better than existing models on the restaurant domain.

Copied to clipboard

Xinyu Zhang, Linmei Hu, Luhao Zhang, Wentao Cheng, Yashen Wang, Ge Shi, Chong Feng, Liqiang Nie

Answer-Supervised Question Reformulation for Enhancing Conversational Machine Comprehension (D19-58)

Challenge:	Existing approaches to optimize sequential recommendation systems rely on item ID sequences, but they lack collaborative knowledge and limited controllability.
Approach:	They propose a simple bi-tuning framework with collaborative information for controllable Large Language Model-based Sequential Recommendation (Laser) they incorporate learnable virtual tokens at prefix and suffix of input text to adapt LLMs with collaborative knowledge .
Outcome:	The proposed framework outperforms state-of-the-art recommendations on real-world datasets.

Copied to clipboard

Qian Li, Hui Su, Cheng Niu, Daling Wang, Zekang Li, Shi Feng, Yifei Zhang

Mitigating the Discrepancy Between Video and Text Temporal Sequences: A Time-Perception Enhanced Video Grounding method for LLM (2025.coling-main)

Challenge:	Existing question reformulation models are based on supervised question labels without considering feedback information from answers.
Approach:	They propose a question reformulation model that integrates conversational history information with reinforcement learning.
Outcome:	The proposed model is more effective in conversational machine comprehension with reinforcement learning.

Copied to clipboard

Xuefen Li, Bo Wang, Ge Shi, Chong Feng, Jiahao Teng

Dolphin: Document Image Parsing via Heterogeneous Anchor Prompting (2025.findings-acl)

Challenge:	Existing video LLMs excel at capturing the overall description of a video but lack the ability to demonstrate an understanding of temporal dynamics and localized content within the video.
Approach:	They propose a Time-Perception Enhanced Video Grounding via Boundary Perception and Temporal Reasoning to improve LLMs' understanding of video temporality.
Outcome:	The proposed method improves on three datasets: ActivityNet, Charades, and DiDeMo (up to 11.2% improvement on R@0.3).

Copied to clipboard

Hao Feng, Shu Wei, Xiang Fei, Wei Shi, Yingdong Han, Lei Liao, Jinghui Lu, Binghong Wu, Qi Liu, Chunhui Lin, Jingqun Tang, Hao Liu, Can Huang

HoLLMwood: Unleashing the Creativity of Large Language Models in Screenwriting via Role Playing (2024.findings-emnlp)

Challenge:	Current document image parsing solutions rely on specialized models or generate content autoregressively.
Approach:	They propose a multimodal document image parsing model that integrates specialized models with autogeneous content generation.
Outcome:	The proposed model achieves state-of-the-art performance across diverse page-level and element-level settings while ensuring superior efficiency.

Copied to clipboard

Jing Chen, Xinyu Zhu, Cheng Yang, Chufan Shi, Yadong Xi, Yuxiang Zhang, Junjie Wang, Jiashu Pu, Tian Feng, Yujiu Yang, Rongsheng Zhang

CGBridge: Bridging Code Graphs and Large Language Models for Better Structure-Aware Code Understanding (2026.findings-acl)

Challenge:	Generative AI has demonstrated unprecedented creativity in the field of computer vision, yet such phenomena have not been observed in the realm of literary creation.
Approach:	They propose a framework for unleashing the creativity of large language models (LLMs) they assign LLMs to different roles involved in real-world scenario, they write .
Outcome:	The proposed framework outperforms baselines in terms of coherence, relevance, interestingness and overall quality on automatically generated screenplays.

Copied to clipboard

Zeqi Chen, Zhaoyang Chu, Yi Gui, Feng Guo, Yao Wan, Chuan Shi

RATION: Entropy-Driven Task-Adaptive Visual Attention Allocation Framework for Multimodal Reasoning (2026.findings-acl)

Challenge:	Existing structure-aware approaches treat structure as serialized text prompts or auxiliary training objectives, failing to provide explicit guidance during inference.
Approach:	They propose a plug-and-play method that enhances Large Language Models with Code Graph information through an external, trainable Bridge module.
Outcome:	The proposed method decouples structural reasoning from textual generation without updating the backbone.

Copied to clipboard

Xingle Xu, Fanheng Kong, Dexian Cai, Shi Feng, Xiaocui Yang, Daling Wang, Yifei Zhang

BERT-BC: A Unified Alignment and Interaction Model over Hierarchical BERT for Response Selection (2024.lrec-main)

Challenge:	Prior studies have focused on strengthening multimodal reasoning by improving representation alignment or increasing computation, but these methods do not characterize the differences in visual demands across tasks.
Approach:	They propose an entropy-driven task-adaptive visual attention allocation framework that uses visual attention entropic as a control signal to dynamically allocate attention according to task demands.
Outcome:	The proposed framework achieves consistent performance gains across diverse reasoning tasks, datasets, and models, providing a clear direction toward more reliable multimodal reasoning.

Copied to clipboard

Zhenfei Yang, Beiming Yu, Yuan Cui, Shi Feng, Daling Wang, Yifei Zhang

TROJail: Trajectory-Level Optimization for Multi-Turn Large Language Model Jailbreaks with Process Rewards (2026.acl-long)

Challenge:	Recent performance boosting for dialogue response selection task achieved by Cross-Encoder based models is limited and the learned models have poor generalization capability in realistic scenarios.
Approach:	They propose a model that combines the representation-based Bi-Encoder and interaction-based Cross-Encoding to achieve better semantic representation.
Outcome:	The proposed model can achieve state-of-the-art performance on three benchmark datasets for multi-turn response selection.

Copied to clipboard

Xiqiao Xiong, Ouxiang Li, Zhuo Liu, Moxin Li, Wentao Shi, Fengbin Zhu, Qifan Wang, Fuli Feng

Few-shot Joint Multimodal Aspect-Sentiment Analysis Based on Generative Multimodal Prompt (2023.findings-acl)

Challenge:	Existing approaches to training multi-turn attackers to probe model safety vulnerabilities rely on turn-level optimization, which is insufficient for learning long-term attack strategies.
Approach:	They propose a multi-turn reinforcement learning problem that optimizes the harmfulness of the final-turn response as the outcome reward.
Outcome:	The proposed approach improves attack success rates across multiple models and benchmarks, highlighting the effectiveness of the proposed approach.

Copied to clipboard

Xiaocui Yang, Shi Feng, Daling Wang, Qi Sun, Wenfang Wu, Yifei Zhang, Pengfei Hong, Soujanya Poria

Sentiment Analysis

Challenge:	Existing studies require massive labeled data to train models for multimodal data analysis.
Approach:	They propose a novel multimodal prompt model that captures specific aspect terms in a few-shot scenario.
Outcome:	The proposed model outperforms baselines on two MABSA-related tasks on a few-shot dataset.

ES4R: Speech Encoding Based on Prepositive Affective Modeling for Empathetic Response Generation (2026.acl-long)

Copied to clipboard

Zhuoyue Gao, Xiaohui Wang, Xiaocui Yang, Wen Zhang, Daling Wang, Shi Feng, Yifei Zhang