Papers by Seung-won Hwang
Two Examples are Better than One: Context Regularization for Gradient-based Prompt Tuning (2023.findings-acl)
Copied to clipboard
| Challenge: | Prompting has gained tremendous attention as an efficient method for the adaptation of large-scale language models. |
| Approach: | They propose a regularization method that guides a prompt to produce a task context properly. |
| Outcome: | The proposed method improves prediction performance in a zero-shot in-context learning setting without demonstration examples for in-constitu learning. |
RaDA: Retrieval-augmented Web Agent Planning with LLMs (2024.findings-acl)
Copied to clipboard
| Challenge: | Agents powered by large language models inherit important limitations such as the restricted context length, dependency on human-engineered exemplars, and insufficient generalization. |
| Approach: | They propose a novel planning method for Web agents that disentangles planning into two stages: for a new given task, it decomposes tasks into high-level subtasks; and then iteratively synthesizes actions based on dynamically retrieved exemplars. |
| Outcome: | The proposed method decomposes tasks into high-level subtasks and iteratively synthesizes actions based on dynamically retrieved exemplars. |
Semi-supervised Training Data Generation for Multilingual Question Answering (L18-1)
Copied to clipboard
| Challenge: | Existing datasets for question answering (QA) tasks mostly support only English . however, existing resources for these tasks are labor intensive . |
| Approach: | They propose to combine Korean QA datasets with machine-translated English resources to build seed resources. |
| Outcome: | The proposed approach leads to 71.50 F1 on Korean QA (comparable to 77.3 F1) |
Query-focused Referentiability Learning for Zero-shot Retrieval (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing dense representations have shown limitations in zero-shot scenarios . however, passage representations fail to align with their gold queries . |
| Approach: | They propose a query-focused concept of 'referentiable' which ensures passage representations are referenced by their gold queries. |
| Outcome: | The proposed model outperforms existing models on the BEIR benchmark. |
PLM-based World Models for Text-based Games (2022.emnlp-main)
Copied to clipboard
| Challenge: | a new study shows that pre-trained world models provide a strong base for world models . worldformer is a text-based game environment that can be used to learn world models in text-driven games. |
| Approach: | They propose to use pre-trained language models to build world models in text-based game environments. |
| Outcome: | The proposed model outperforms state-of-the-art model-free algorithms in Atari games while retaining sample efficiency. |
NL2pSQL: Generating Pseudo-SQL Queries from Under-Specified Natural Language Questions (D19-1)
Copied to clipboard
| Challenge: | Existing studies focus on generating SQL codes from natural language questions . however, questions cover more diverse tasks including table manipulation or performance issues . |
| Approach: | They propose a task to generate pSQL codes from natural language questions . they define two new metrics suitable for the task, Canonical-BLEU and SQL-BLUE . |
| Outcome: | The proposed task generates well-formed queries on under-specified database issues. |
Chaining Event Spans for Temporal Relation Grounding (2024.eacl-long)
Copied to clipboard
| Challenge: | Existing approaches to understanding temporal relations between events have relied on answer overlaps as a proxy label to distinguish similar and dissimilar questions. |
| Approach: | They propose a timeline reasoning network that elicits proper reasoning behaviors through a module for predicting time spans of events. |
| Outcome: | The proposed approach outperforms existing methods by resolving spurious overlaps using the predicted timeline. |
ArchCode: Incorporating Software Requirements in Code Generation with Large Language Models (2024.acl-long)
Copied to clipboard
| Challenge: | Despite the critical role of software requirements, these criteria have not been studied actively in previous code generation works. |
| Approach: | They propose a framework that leverages in-context learning to organize and extrapolate unexpressed requirements from textual descriptions. |
| Outcome: | The proposed framework generates functional requirements from textual descriptions and extrapolates unexpressed requirements from them. |
BotsTalk: Machine-sourced Framework for Automatic Curation of Large-scale Multi-skill Dialogue Datasets (2022.emnlp-main)
Copied to clipboard
| Challenge: | a number of largescale datasets targeting a specific conversational skill have recently become available. |
| Approach: | They propose a framework where multiple agents grounded to specific skills participate in a conversation to automatically annotate multi-skill dialogues. |
| Outcome: | The proposed framework can be used to build open-domain chatbots with diverse communicative skills. |
D3: Dynamic Docid Decoding for Multi-Intent Generative Retrieval (2026.eacl-industry)
Copied to clipboard
| Challenge: | Existing GR systems rely on offline DocID assignment and constrained decoding . offline Doc ID assignment and decoding often prevents GR from capturing query-specific intent . |
| Approach: | They propose a mechanism that adaptively refines DocIDs through query-informed identifier expansion. |
| Outcome: | The proposed mechanism improves retrieval accuracy on unseen and multi-intent documents. |
Plug-and-Play Adaptation for Continuously-updated QA (2022.findings-acl)
Copied to clipboard
| Challenge: | Existing tasks to assess LMs’ efficacy as KBs do not adequately consider multiple large-scale updates. |
| Approach: | They propose a task where multiple large-scale updates are made to language models and plug-in modules are used to handle the updates. |
| Outcome: | The proposed method outperforms existing methods on zsRE QA and NQ datasets and is 4x more effective in terms of updates/forgets ratio compared to a fine-tuning baseline. |
Modularized Transfer Learning with Multiple Knowledge Graphs for Zero-shot Commonsense Reasoning (2022.naacl-main)
Copied to clipboard
| Challenge: | Currently, commonsense reasoning systems are limited by expensive data annotations and overfitting to a specific benchmark. |
| Approach: | They propose to transform a commonsense knowledge graph into synthetic QA-form samples for model training. |
| Outcome: | The proposed framework improves performance with multiple commonsense KGs on five commonsensense reasoning benchmarks. |
GRAD: Generalizing RAG Adaptation with Decoding (2026.acl-long)
Copied to clipboard
| Challenge: | Using GRAD, we can steer Retrieval-augmented generation objectives without retraining large language models. |
| Approach: | They propose an adaptive decoding-time framework that keeps the base generator fixed and composes small, objective-specific guidance at inference. |
| Outcome: | The proposed framework improves accuracy with favorable latency across public benchmarks and private settings with no in-domain labels while reliably activating helpful objectives and suppressing harmful ones, adaptively to tasks. |
ReACC: A Retrieval-Augmented Code Completion Framework (2022.acl-long)
Copied to clipboard
| Challenge: | Recent work has shown that statistical language modeling with transformers can greatly improve the performance in code completion tasks. |
| Approach: | They propose a retrieval-augmented code completion framework that combines a source code retriever and an auto-regressive language model for programming language. |
| Outcome: | The proposed framework achieves state-of-the-art on CodeXGLUE benchmark. |
On Complementarity Objectives for Hybrid Retrieval (2023.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to hybrid retrieval focus on sparse models to capture “residual” features neglected in spars. |
| Approach: | They propose a new objective to capture a fuller notion of complementarity . they propose to improve the model's Ratio of Complementarity to improve RoC . |
| Outcome: | The proposed method outperforms state-of-the-art methods on three representative IR benchmarks with statistical significance. |
HARP: Hesitation-Aware Reframing in Transformer Inference Pass (2025.naacl-long)
Copied to clipboard
| Challenge: | a recent study has shown that inference steps are not equally challenging, with some being "harder" and others "easier." |
| Approach: | They propose a modified Transformer forward pass that selectively applies additional computation when the model encounters uncertainty during token generation. |
| Outcome: | The proposed method achieves performance gains while maintaining inference times twice faster than beam search. |
Query Variant Detection Using Retriever as Environment (2025.naacl-industry)
Copied to clipboard
| Challenge: | Identifying query variants is nontrivial as highly similar query pairs may fail to differ in word form, order, or phrasing despite sharing the same intent. |
| Approach: | They propose to use retrieval as an environment feedback to capture semantic equivalence . experimental results demonstrate the efficacy of the proposed method . |
| Outcome: | The proposed method improves query variant detection across diverse scenarios. |
Soft Representation Learning for Sparse Transfer (P19-1)
Copied to clipboard
| Challenge: | Using adversarial training, we can “soft-code” shared and private spaces to avoid sparse sharing. |
| Approach: | They propose to use adversarial training to “soft-code” shared and private spaces to avoid the shared space gets too sparse. |
| Outcome: | The proposed architecture avoids sparse sharing of shared and private spaces, and also deals with low-quality input. |
DADA: Distribution-Aware Domain Adaptation of PLMs for Information Retrieval (2024.findings-acl)
Copied to clipboard
| Challenge: | Pre-trained language models struggle with out-of-domain data due to distribution shifts . generative domain adaptation (DA) methods have been proposed to address these issues . |
| Approach: | They propose a distribution-aware domain adaptation method to address distribution shifts in domains . they use observation-level feedback and observation- level feedback to adapt to the target domain . |
| Outcome: | The proposed method adapts to the domain distribution knowledge at the level of a single document and the corpus and expands document representation to unseen gold query terms using domain and observation feedback. |
Retrieval-Augmented Controllable Review Generation (2020.coling-main)
Copied to clipboard
| Challenge: | Existing approaches to generate reviews using attribute identifiers are limited and dependent on how well they can capture vector representations of attributes. |
| Approach: | They propose to leverage attributes as inputs for review generation by using reference sets . they propose to use these references to enrich inductive biases of given attributes . |
| Outcome: | The proposed model improves over previous approaches on automatic and human evaluation metrics. |
Query Generation for Multimodal Documents (2021.eacl-main)
Copied to clipboard
| Challenge: | Existing approaches to find relevance for multimodal documents with images are expensive and require a lot of runtime overhead. |
| Approach: | They propose to attach generated queries to doc-uments and index them to narrow down to candidate matches using inverted index. |
| Outcome: | The proposed model improves relevance ranking for multimodal documents with images . the proposed model can achieve the state of the art in the first stage retrieval scenarios . |
TAGQuant: Token-Aware Clustering for Group-Wise Quantization (2026.eacl-industry)
Copied to clipboard
| Challenge: | Existing work clusters channels using token dimension, which is suboptimal for grouping . a common challenge in LLM quantization is supporting "group-wise" quantization . |
| Approach: | They propose a method to group channels with similar activation distributions using tokens . they propose shuffle operation that reduces relative GSM8K error by 86% . |
| Outcome: | The proposed method reduces GSM8K error by 86% in both INT4 and MXFP4 formats compared to baselines . |
Mining Cross-Cultural Differences and Similarities in Social Media (P18-1)
Copied to clipboard
| Challenge: | a new paper examines the problem of computing cross-cultural differences and similarities in natural language understanding . cross-culture differences are important for cross-lingual research, especially in social media . |
| Approach: | They propose a framework for computing cross-cultural differences and similarities from social media . they propose to use a social media platform to find similar terms for slang across languages . |
| Outcome: | The proposed framework outperforms baseline methods on two novel tasks. |
Debiasing Event Understanding for Visual Commonsense Tasks (2022.findings-acl)
Copied to clipboard
| Challenge: | a recent study shows that object-based event understanding is purely likelihood-based, leading to incorrect event prediction. |
| Approach: | They propose to mitigate object-based event understanding by optimizing aggregation with association-based prediction. |
| Outcome: | The proposed approach improves visual commonsense reasoning tasks by combining do-calculus with association-based prediction. |
HIL: Hybrid Isotropy Learning for Zero-shot Performance in Dense retrieval (2024.naacl-long)
Copied to clipboard
| Challenge: | Recent advances in dense retrieval models have brought ColBERT to prominence in information retrieval, but it is underperforming in zero-shot tasks. |
| Approach: | They propose a Hybrid Isotropy Learning architecture that integrates isotropic and anisotropic representations to improve zero-shot retrieval performance. |
| Outcome: | The proposed model outperforms the baseline ColBERT model in BEIR benchmarks. |
R3-SQL: Ranking Reward and Resampling for Text-to-SQL (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing rankers assign inconsistent scores to functionally equivalent SQL queries . ranking cannot recover when the correct SQL is absent from the pool. |
| Approach: | They propose a Text-to-SQL framework that rewards ranking and resampling . it first groups candidates by execution result and ranks groups for consistency . |
| Outcome: | The proposed framework achieves 75.03 execution accuracy on BIRD-dev, a new state of the art among methods using models with disclosed sizes. |
Evaluating Research Novelty Detection: Counterfactual Approaches (D19-53)
Copied to clipboard
| Challenge: | Despite its importance, this direction of research has not been explored as much. |
| Approach: | They propose to use counterfactual simulations to evaluate paper novelty detection models . they ask models to differentiate papers at time t and counterf actual paper from future time . |
| Outcome: | The proposed models can be compared against a set of papers with a given date and with different annotations. |
ScriptMix: Mixing Scripts for Low-resource Language Parsing (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing work has considered transliteration and vocabulary augmentation, but the consideration of combining the two has been lacking. |
| Approach: | They propose a multilingual pretrained language model that combines two strengths and overcomes the hurdle of combining them. |
| Outcome: | The proposed model improves POS accuracy by 14% and improves DEP LAS score by 5.6%. |
Breaking ReLU Barrier: Generalized MoEfication for Dense Pretrained Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to convert pretrained dense models to MoEs are limited to ReLU-based models with natural sparsity. |
| Approach: | They propose a G-MoEfication approach for arbitrary dense models where activation sparsity assumptions no longer hold. |
| Outcome: | The proposed method reduces the inference cost associated with dense models by sparsely activating experts. |
Retrieval-augmented Video Encoding for Instructional Captioning (2023.findings-acl)
Copied to clipboard
| Challenge: | Instructional videos provide a detailed multimodal context of each procedure in instruction. key-object degeneracy is a problem for machine systems, causing incorrect captions. |
| Approach: | They propose a retrieval-based framework to augment the model representations in the presence of key-object degeneracy. |
| Outcome: | The proposed framework can be extended over baselines using modalities with key-object degeneracy. |
CORD: Balancing COnsistency and Rank Distillation for Robust Retrieval-Augmented Generation (2025.naacl-short)
Copied to clipboard
| Challenge: | Existing methods to ground large language models fail to adequately attend to all contexts . position bias is hindered by retrieval-augmented generation, which requires constant attention . |
| Approach: | They propose to augment and distill training instances with their perturbed positions to encourage consistent predictions . they also propose to balance COnsistency and Rank Distillation by combining noise-controlled perturbations with augmentation and distillation. |
| Outcome: | The proposed method outperforms existing methods in diverse RAG benchmarks. |
Beyond Markovian Forgetfulness: Episodic Memory for Reasoning-Intensive Retrieval (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for reasoning-intensive information retrieval suffer from inefficiency . Chain-of-Thought (CoT) approaches suffer from lack of token efficiency . Existing models lack episodic memory, which stores the history of prior states . |
| Approach: | They propose an algorithm that enhances state-based frameworks with an episodic memory module that stores the full history of prior states for a query. |
| Outcome: | The proposed model outperforms CoT and state-based models on the BRIGHT benchmark and is highly token-efficient. |
Learning to Rank Generation with Pairwise Partial Rewards (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for conditional text generation suffer from large action space and delayed reward, as the reward can be computed only after an entire sequence is generated. |
| Approach: | They propose a method that provides partial rewards for intermediate actions taken on partial sequences to prioritize actions that lead to the generation of more desirable sequences. |
| Outcome: | The proposed method overcomes the limitations of the prevalent supervised maximum likelihood estimation approach. |
Privacy-Preserving Text Classification on BERT Embeddings with Homomorphic Encryption (2022.naacl-main)
Copied to clipboard
| Challenge: | Embeddings compress information into low-dimensional vectors, but can leak private information about sensitive attributes of text. |
| Approach: | They propose a method to privatize embeddings based on homomorphic encryption to prevent leakage of sensitive information in the process of text classification. |
| Outcome: | The proposed method can protect embeddings from leakage while preserving their utility on downstream tasks. |
CR-COPEC: Causal Rationale of Corporate Performance Changes to learn from Financial Reports (2023.findings-emnlp)
Copied to clipboard
Ye Chun, Sunjae Kwon, Kyunghwan Sohn, Nakwon Sung, Junyoup Lee, Byoung Seo, Kevin Compher, Seung-won Hwang, Jaesik Choi
| Challenge: | CR-COPEC is a large-scale domain-adaptation causal sentence dataset to detect financial performance changes of corporate. |
| Approach: | They propose a large-scale domain-adaptation causal sentence dataset to detect financial performance changes of corporate. |
| Outcome: | The proposed dataset can be used by investors and analysts without having to read through all the documents. |
COMMIT: Code-Mixing English-Centric Large Language Model for Multilingual Instruction Tuning (2024.findings-naacl)
Copied to clipboard
| Challenge: | COMMIT improves the exact match score of low-resource language QA by 32x. |
| Approach: | They propose to specialize instruction tuning to deviate from English-centric instruction tuning . they propose to perform cross-lingual alignment to overcome data imbalance . |
| Outcome: | The proposed method improves the exact match score of low-resource language QA by 32x. |
Visual Choice of Plausible Alternatives: An Evaluation of Image-based Commonsense Causal Reasoning (L18-1)
Copied to clipboard
Jinyoung Yeo, Gyeongbok Lee, Gengyu Wang, Seungtaek Choi, Hyunsouk Cho, Reinald Kim Amplayo, Seung-won Hwang
| Challenge: | Existing methods for evaluating plausibility of events are focused on measuring causal dependency between events or actions. |
| Approach: | They propose a task to identify the more plausible alternative with their commonsense causal context. |
| Outcome: | The proposed task is based on a visual COPA dataset with 380 questions and over 1K images with various topics. |
Interventional Speech Noise Injection for ASR Generalizable Spoken Language Understanding (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to increase the robustness of pre-trained language models (PLMs) against unseen ASR systems produce noisy inputs for SLU models, which can significantly degrade their performance. |
| Approach: | They propose to introduce ASR-plausible noises into pre-trained language models by cutting off the non-causal effect of noises. |
| Outcome: | The proposed method improves the robustness and generalizability of SLU models against unseen ASR systems by cutting off the non-causal effect of noises. |
MICRON: Multigranular Interaction for Contextualizing RepresentatiON in Non-factoid Question Answering (D19-1)
Copied to clipboard
| Challenge: | Existing approaches for non-factoid question answering can be categorized into representation and interaction focused approaches. |
| Approach: | They propose a novel approach which derives contextualized uni-gram representation from n-grams. |
| Outcome: | The proposed approach achieves state-of-the-art in two public non-factoid question answering datasets. |
Agentic Verification for Ambiguous Query Disambiguation (2026.findings-acl)
Copied to clipboard
Youngwon Lee, Seung-won Hwang, Ruofan Wu, Feng Yan, Danmei Xu, Moutasem Akkad, Zhewei Yao, Yuxiong He
| Challenge: | Prior Diversify-then-Verify pipelines generate interpretations and then retrieve evidence . ambiguous queries require RAG to disambiguate into interpretations that can be answered from corpus . |
| Approach: | They propose a novel approach that unifies diversification with verification by integrating retriever relevance and generator answerability feedback early. |
| Outcome: | The proposed approach improves grounding-aware F1 by 23% over baselines across multiple LLMs. |
Multilingual Lottery Tickets to Pretrain Language Models (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Multilingual pretrained language models (mPLMs) have become the de-facto standard for multilingual tasks. |
| Approach: | They propose to search for a per-language subnetwork with comparable performance to the full model by scaling the model to reduce interference and then redistributing parameters to keep the parameters reduced. |
| Outcome: | The proposed model reduces the inference cost of models for each language while keeping the capacity per language more or less the same. |
Relevance-assisted Generation for Robust Zero-shot Retrieval (2023.emnlp-industry)
Copied to clipboard
| Challenge: | Despite strong in-domain performance, dense retrievers have shown poor generalization to out-of-domain zero-shot tasks where no training queries are available. |
| Approach: | They propose to generate domain-specific pseudo queries for fine-tuning with domain-relevant relevance between PQ and documents. |
| Outcome: | The proposed approach is more robust to domain shifts, validated on BEIR zero-shot tasks. |
SQuAD2-CR: Semi-supervised Annotation for Cause and Rationales for Unanswerability in SQuAD 2.0 (2020.lrec-1)
Copied to clipboard
| Challenge: | Existing models are brittle for adversarial perturbed questions, causing uncertainty . a dataset with annotations on unanswerable questions is not available to solve this problem . |
| Approach: | They use crowdsourced annotations to annotate unanswerable questions . they also annotated which part of the question causes unanswered questions a . |
| Outcome: | The proposed dataset can be used to improve model interpretation, authors say . they find that existing models are brittle for adversarial perturbed questions . |
Relevance to Utility: Process-Supervised Rewrite for RAG (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing bridge modules attempt to rewrite documents for better generation, but fail to capture "document utility". |
| Approach: | They propose to observe reasoning process as process supervision and scale this observation to enhance reliability in distillation. |
| Outcome: | The proposed method improves over strong bridging baselines on multiple open-domain question-answering benchmarks. |
Less is More: Attention Supervision with Counterfactuals for Text Classification (2020.emnlp-main)
Copied to clipboard
| Challenge: | Specifically, we explore the advantage of counterfactual reasoning, over associative reasoning . Adding human supervision to attention has been shown to improve model predictions and explanations . |
| Approach: | They propose to use machine-augmented human attention supervision to enhance model quality. |
| Outcome: | The proposed method is more effective than existing methods requiring higher annotation cost . the proposed method can be trained to generate similar attention to human supervision . |
Structure-Augmented Keyphrase Generation (2021.emnlp-main)
Copied to clipboard
| Challenge: | Creating keyphrases that are likely to be words absent from the given document is challenging . |
| Approach: | They propose novel keyphrase generation tasks that augment missing context by adding keyphrases to documents. |
| Outcome: | The proposed keyphrase generation task outperforms the state-of-the-art in two keyphrase tasks. |
On Sample-Efficient Code Generation (2023.emnlp-industry)
Copied to clipboard
Hojae Han, Yu Jin Kim, Byoungjip Kim, Youngwon Lee, Kyungjae Lee, Kyungmin Lee, Moontae Lee, Kyunghoon Bae, Seung-won Hwang
| Challenge: | Existing approaches to code generation rely on rejection sampling to generate multiple code snippets then select the best. |
| Approach: | They propose a framework that prioritizes sampling on test problems that models can solve. |
| Outcome: | The proposed framework reduces sampling costs while maintaining comparable code generation performance. |
PROM: Pivoted and Regulated Optimization for Multilingual Instruction Learning (2025.naacl-short)
Copied to clipboard
| Challenge: | Existing solutions to large language models (LLMs) are English-centric, hindering their application to 6500+ existing languages. |
| Approach: | They propose to append English tuning data with its translated pair to solve this problem . they identify English as an internal pivot language and propose to regulate between them . |
| Outcome: | The proposed model is able to generalize on multiple benchmarks across different languages. |
Counterfactual-Consistency Prompting for Relative Temporal Understanding in Large Language Models (2025.acl-short)
Copied to clipboard
| Challenge: | Existing work has highlighted that large language models lack temporal reasoning abilities, especially when attempting to infer temporal relationships without relying on absolute time indicators. |
| Approach: | They propose a method that generates counterfactual questions and enforces collective constraints, enhancing the model’s consistency. |
| Outcome: | The proposed method shows significant improvements in event ordering for explicit and implicit events and temporal commonsense understanding. |
Normalizing Mutual Information for Robust Adaptive Training for Translation (2022.emnlp-main)
Copied to clipboard
| Challenge: | Neural machine translation models have been reported to generate hallucinations . despite the success of the models, there are still challenges to improve fluency . |
| Approach: | They propose a scoring metric for the importance of target sentences and tokens to encourage fluent translations. |
| Outcome: | The proposed metric improves translation fluency and source-faithfulness . the proposed nmi model is not properly normalized, the authors argue . |
UnIte: Uncertainty-based Iterative Document Sampling for Domain Adaptation in Information Retrieval (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods focus on diversity but fail to capture model uncertainty. |
| Approach: | They propose a method to generalize neural retrievers to an unseen domain by generating pseudo queries on target domain documents. |
| Outcome: | The proposed method improves performance on large datasets with small and large models while limiting the learning utility of the current model. |
Coffee-Gym: An Environment for Evaluating and Improving Natural Language Feedback on Erroneous Code (2024.emnlp-main)
Copied to clipboard
Hyungjoo Chae, Taeyoon Kwon, Seungjun Moon, Yongho Song, Dongjin Kang, Kai Ong, Beong-woo Kwak, Seonghyeon Bae, Seung-won Hwang, Jinyoung Yeo
| Challenge: | Large language models (LLMs) have made great progress in code generation, however, they still produce errors. |
| Approach: | They propose a RL environment that provides feedback on code editing by analyzing the performance of the revised code in unit tests. |
| Outcome: | The proposed model outperforms baselines in enhancing open-source code LLMs’ code editing, making them comparable with closed-source LLM. |
ContrastiveMix: Overcoming Code-Mixing Dilemma in Cross-Lingual Transfer for Information Retrieval (2024.naacl-short)
Copied to clipboard
| Challenge: | Multilingual pretrained language models have been widely adopted in cross-lingual transfer . however, training mPLMs on code-mixed data is counterproductive . |
| Approach: | They propose a method that disentangles contrastive loss between conflicting objectives . they propose mPLMs that train on code-mixed data for information retrieval . |
| Outcome: | The proposed model outperforms mDPR in cross-lingual transfer tasks . it is effective in sentence-level and token-level zero-shot tasks despite lack of target language data . |
FaVe: Factored and Verified Search Rationale for Long-form Answer (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing solutions for long-form question-answering (LFQA) use chain-of-thought (CoT) with retrieval-augmented generation (RAG). |
| Approach: | They propose to integrate chain-of-thought (CoQ) with retrieval-augmented generation to improve answer comprehensiveness and verifiability. |
| Outcome: | The proposed approach outperforms ChatGPT baselines while maintaining efficiency. |
Disentangling Questions from Query Generation for Task-Adaptive Retrieval (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing work generates synthetic queries from domain-specific documents to jointly train the retriever. |
| Approach: | They propose a query generator that better adapts to wide search intents expressed in the BeIR benchmark. |
| Outcome: | The proposed query generator outperforms baselines and existing models on tasks with underexplored intents while using a query generator 47 times smaller than the previous state-of-the-art. |
From Token to Action: State Machine Reasoning to Mitigate Overthinking in Information Retrieval (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Chain-of-Thought (CoT) prompting often leads to overthinking in large language models . redundant trajectories that revisit similar states and misguided reasoning that diverges from user intent are two key challenges in information retrieval. |
| Approach: | They propose a transition-based reasoning framework that supports early stopping and fine-grained control. |
| Outcome: | The proposed framework improves retrieval performance by 3.4% while reducing token usage by 74.4%. |
Inference Scaling for Bridging Retrieval and Augmented Generation (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing work observed the generator bias, such that improving the retrieval results may negatively affect the outcome. |
| Approach: | They propose to use inference scaling to aggregate inference calls from the permuted order of retrieved contexts to create a new ranking. |
| Outcome: | The proposed approach improves ROUGE-L on MS MARCO and EM on HotpotQA benchmarks by 7 points. |
ECoRAG: Evidentiality-guided Compression for Long Context RAG (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing compression methods do not focus on filtering out non-evidential information, which limit performance in LLMs. |
| Approach: | They propose a framework that leverages external documents through Retrieval-Augmented Generation (RAG) to reduce RAG overhead. |
| Outcome: | Experiments show that ECoRAG outperforms existing compression methods on ODQA tasks. |
Entity Commonsense Representation for Neural Abstractive Summarization (N18-1)
Copied to clipboard
| Challenge: | Current ELS’s are not sufficiently effective, possibly introducing unresolved ambiguities and irrelevant entities. |
| Approach: | They propose an off-the-shelf entity linking system to extract linked entities and propose Entity2Topic (E2T) module attachable to a sequence-to-sequence model that transforms a list of entities into a vector representation of the topic of the summary. |
| Outcome: | The proposed model improves the performance of the Gigaword and CNN summarization datasets by at least 2 ROUGE points. |
Intervention-Based Alignment of Code Search with Execution Feedback (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing code search training datasets approximate text-code co-occurrences as positive execution feedback, but this approximation may misalign models’ retrieval decisions from ground-truth correctness. |
| Approach: | They propose a code intervention-based reinforcement learning approach that perturbs training code to result in misalignment, then tests models’ decisions and corrects them with the execution feedback by reinforcement learning. |
| Outcome: | The proposed method induces the execution feedback from perturbation, without actual execution, and then tests models’ decisions and corrects them with the execution input by reinforcement learning. |
Towards Compositional Generalization in Code Search (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing code search models that focus on code as an unstructured sequence fail to generalize when the lexical perturbation without changing structures and labels is applied in test codes. |
| Approach: | They propose a compositional generalization model that extracts structural elements and a code template that targets compositional genericization. |
| Outcome: | The proposed model is complementary to flow graphs in GraphCodeBERT, by enhancing structural context around variables. |
tRAG: Term-level Retrieval-Augmented Generation for Domain-Adaptive Retrieval (2025.naacl-long)
Copied to clipboard
| Challenge: | Neural retrieval models suffer when there is a domain shift between training and test data distributions. |
| Approach: | They propose to generate domain-adapted pseudo-queries using large language models (LLMs) to improve term recall of unseen query terms by using term-level Retrieval-Augmented Generation (tRAG). |
| Outcome: | The proposed method significantly improves recall for unseen terms by 10.6% and outperforms LLM and retrieval-augmented generation baselines on overall retrieval performance. |
Evidentiality-aware Retrieval for Overcoming Abstractiveness in Open-Domain Question Answering (2024.findings-eacl)
Copied to clipboard
| Challenge: | Existing approaches to ODQA use a simple yet effective retriever-reader framework, but this approach is not always effective in abstractive tasks. |
| Approach: | They propose a method that leverages synthetic distractor samples to learn to discriminate evidence passages from distractors. |
| Outcome: | The proposed method is validated on multiple abstractive open-domain question answering tasks. |
Agent-as-Judge for Factual Summarization of Long Narratives (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have demonstrated near-human performance in summarization tasks based on traditional metrics such as ROUGE and BERTScore . however, these metrics do not adequately capture critical aspects of summarizing quality, such as factual accuracy, especially for long narratives. |
| Approach: | They propose a framework that evaluates and refines factuality in narrative summarization by leveraging a Character Knowledge Graph extracted from input narrative. |
| Outcome: | The proposed framework evaluates factuality and provides actionable guidance for refinement. |
RoToR: Towards More Reliable Responses for Order-Invariant Inputs (2025.acl-long)
Copied to clipboard
| Challenge: | Existing solutions to positional bias in listwise inputs are limited on practical problems . e.g., lost-in-the-middle problem is a common problem for listwise models . |
| Approach: | They propose a zero-shot order-invariant LM for order- invariant inputs with minimal modifications of positional IDs and Selective Routing for listwise tasks. |
| Outcome: | The proposed framework can handle order-invariant and sensitive inputs in listwise tasks. |
STUN: Structured-Then-Unstructured Pruning for Scalable MoE Pruning (2025.acl-long)
Copied to clipboard
| Challenge: | Mixture-of-experts (MoEs) have been adopted for reducing inference costs by sparsely activating experts in large language models (LLMs). |
| Approach: | They propose a structured-then-unstructured approach outperforming both of structured and unstructured pruning for MoEs. |
| Outcome: | The proposed approach outperforms both of structured and unstructured pruning, especially for MoEs with hundreds of experts. |
Collective Relevance Labeling for Passage Retrieval (2022.naacl-main)
Copied to clipboard
| Challenge: | Existing approaches to improve IR labels are incomplete and require computational overheads. |
| Approach: | They propose to distill knowledge for informed labeling without high computation overheads at evaluation time. |
| Outcome: | The proposed model outperforms state-of-the-art models while distilling the rankings better. |
Consistency is Key: On Data-Efficient Modality Transfer in Speech Translation (2023.findings-emnlp)
Copied to clipboard
| Challenge: | End-to-end approaches to speech translation suffer from data scarcity compared to machine translation (MT). |
| Approach: | They propose a method which combines knowledge distillation and consistency learning to break the dilemma of learning-forgetting. |
| Outcome: | The proposed method outperforms the previous methods on a MuST-C dataset even without additional data. |
Learning with Limited Data for Multilingual Reading Comprehension (D19-1)
Copied to clipboard
| Challenge: | Existing approaches to support question answering in a new language with limited training resources introduce noises to the training data due to translation or generation errors. |
| Approach: | They propose a weakly-supervised framework that quantifies noises from automatically generated labels to deemphasize or fix noisy data in training. |
| Outcome: | The proposed framework can deemphasize or fix noisy data in training on low-resource languages with varying similarity to English. |
Cold-Start Aware User and Product Attention for Sentiment Classification (P18-1)
Copied to clipboard
| Challenge: | Existing models do not deal with cold-start problem typical in review websites. |
| Approach: | They propose a Hybrid Contextualized Sentiment Classifier that uses word encoder and Cold-Start Aware Attention to pool word vectors. |
| Outcome: | The proposed model performs significantly better on famous datasets despite having less complexity and can be trained much faster. |
When to Read Documents or QA History: On Unified and Selective Open-domain QA (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing work aims to answer factoid questions from an open set of domains using knowledge sources. |
| Approach: | They propose to use QA-pair and document corpora to answer open-domain questions . they propose to apply natural follow-up to both models to find answers . |
| Outcome: | The proposed method is validated on natural questions and TriviaQA. |
Robustifying Multi-hop QA through Pseudo-Evidentiality Training (2021.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to robustify multi-hop question answering models require expensive annotations. |
| Approach: | They propose a method to supervise answers with right reasoning chains without annotations . they compare answers confidence with and without evidence sentences to generate "pseudo-evidentiality" annotations. |
| Outcome: | The proposed model is accurate and robust in multi-hop reasoning. |
Adaptive Retrieval for Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing reasoning-based rerankers suffer from bounded recall. |
| Approach: | They propose a framework that leverages adaptive retrieval to ensure sufficient "bridge" documents are retrieved for reasoning-intensive retrieval. |
| Outcome: | The proposed method outperforms baselines on reasoning-intensive retrieval tasks by 5.6%pt. |
CoEx – Co-evolving World-model and Exploration (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing LLM agents fail to assimilate new observations into dynamic updates of the world model, leading to divergent and erroneous plans. |
| Approach: | They propose a hierarchical agent architecture that allows LLM planning to co-evolve with a dynamically updated model of the world. |
| Outcome: | The proposed agent outperforms existing agent paradigms in planning and exploration. |
Intended Target Identification for Anomia Patients with Gradient-based Selective Augmentation (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Identifying the intended target item from patient’s circumlocution involves the two challenges of term failure and error. |
| Approach: | They propose to robustify the model from unseen and SPE terms and enhance it with unseense terms by using gradient-based selective augmentation (GradSelect). |
| Outcome: | The proposed model outperforms existing models on the Tip of the Tongue dataset and shows that it can handle anomia patients by addressing the outlined challenges. |
Towards Lifelong Dialogue Agents via Timeline-based Memory Management (2025.naacl-long)
Copied to clipboard
Kai Tzu-iunn Ong, Namyoung Kim, Minju Gwak, Hyungjoo Chae, Taeyoon Kwon, Yohan Jo, Seung-won Hwang, Dongha Lee, Jinyoung Yeo
| Challenge: | Existing studies focus on getting rid of outdated memories to improve retrieval quality, but we argue that such memories provide rich, important contextual cues for response generation (RG). |
| Approach: | They propose a framework for LLM-based lifelong dialogue agents that discards memory removal and manages large-scale memories by linking them based on their temporal and cause-effect relation. |
| Outcome: | The proposed framework augments RG with memory timelines based on evolution or causality of relevant past events. |
Pseudo-Relevance for Enhancing Document Representation (2022.emnlp-main)
Copied to clipboard
| Challenge: | a novel approach to document retrieval can be used to encode documents as vectors . a few query-relevant terms can be pruned out to reduce index overhead . |
| Approach: | They propose to enhance the document representation for the bi-encoder approach in dense document retrieval. |
| Outcome: | The proposed solution reduces latency and memory footprint up to 8- and 3-fold . it is validated on MSMARCO and real-world search query logs . |
DuET: Dual Execution for Test Output Prediction with Generated Code and Pseudocode (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent studies have shown that test output prediction is difficult to achieve due to code errors. |
| Approach: | They propose a framework that grounds prediction on error-resilient pseudocode and simulates execution via LLM reasoning to overcome limitations of direct execution suffering from code errors. |
| Outcome: | The proposed framework improves Pass@1 on LiveCodeBench, BigCodeBech-Hard, DevEval and HumanEval(+) and improves on pass@1 by 13.6 pp. |
ListT5: Listwise Reranking with Fusion-in-Decoder Improves Zero-shot Retrieval (2024.acl-long)
Copied to clipboard
| Challenge: | Existing listwise reranking models rely on pointwise sizing of each passage . Until now, listwise models lack the ability to compare between passages at inference time . |
| Approach: | They propose a listwise reranking approach based on Fusion-in-Decoder that handles multiple candidate passages at train and inference time. |
| Outcome: | The proposed model outperforms the state-of-the-art RankT5 model on the BEIR benchmark for zero-shot retrieval task with a notable +1.3 gain in the average NDCG@10 score. |
Smarter, Not Harder: Training-Free Adaptive Computation for Transformers (2025.findings-acl)
Copied to clipboard
| Challenge: | Adaptive Computation in Transformers (ACT) aims to reduce computation for simpler inferences while enhancing performance by allocating more computation to complex inference steps. |
| Approach: | They propose a method that perturbs network weights rather than input embeddings to improve performance. |
| Outcome: | The proposed method outperforms beam search and hesitation-based methods but suffers from inefficiency and instability due to its reliance on randomness. |
FAD-X: Fusing Adapters for Cross-lingual Transfer to Low-Resource Languages (2022.aacl-short)
Copied to clipboard
| Challenge: | Adapter-based tuning is a technique that selectively updates language-specific parameters to adapt to a new language, rather than fine-tuning all shared weights. |
| Approach: | They propose to add light-weight adapters to multilingual pretrained language models (mPLMs) and add language-specific parameters to adapt to a new language. |
| Outcome: | The proposed adapter can enhance cross-lingual transfer from pretrained adapters for well-known named entity recognition and classification benchmarks. |
PERC: Plan-As-Query Example Retrieval for Underrepresented Code Generation (2025.coling-main)
Copied to clipboard
| Challenge: | Using large language models to generate code has shown significant promise, but selecting effective examples to improve generation quality remains a challenging task. |
| Approach: | They propose a framework that utilizes algorithmic plans to identify and retrieve effective examples. |
| Outcome: | The proposed framework outperforms the state-of-the-art RAG methods in code generation even when the source and target languages match or differ. |
Tree-of-Prompts: Abstracting Control-Flow for Prompt Optimization (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing prompt optimization methods struggle with disjoint cases in complex tasks. |
| Approach: | They propose a tree-of-prompts structure which expands child prompts from parent prompts . they propose to use a nested if-else structure to address varying similarities and complexities . |
| Outcome: | The proposed tree-of-prompts outperforms PromptAgent and MoP on Gorilla, MATH and subset of BBH benchmarks. |
PLEX: Adaptive Parameter-Efficient Fine-Tuning for Code LLMs using Lottery-Tickets (2025.naacl-industry)
Copied to clipboard
| Challenge: | PLEX is a lottery-ticket based parameter-efficient fine-tuning method that adapts large language models to well-supported and underrepresented programming languages (PLs) in pretraining. |
| Approach: | They propose a lottery-ticket based parameter-efficient fine-tuning method that adapts large language models to well-supported and underrepresented programming languages (PLs) |
| Outcome: | The proposed method achieves state-of-the-art performance among PEFT methods while maintaining competitive results with reduced computational overhead. |
Mind the Gap! Injecting Commonsense Knowledge for Abstractive Dialogue Summarization (2022.coling-1)
Copied to clipboard
| Challenge: | Existing frameworks that use commonsense as supervision only use input knowledge, but it generates more informative and consistent summaries. |
| Approach: | They propose to leverage the unique characteristics of dialogues sharing commonsense knowledge to solve the difficulties in summarizing them. |
| Outcome: | The proposed framework generates more informative and consistent summaries with injected commonsense knowledge than existing methods. |
QuBE: Question-based Belief Enhancement for Agentic LLM Reasoning (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in Large Language Models (LLMs) have led to an explosion of interest in their deployment as agents. |
| Approach: | They propose a method that enhances agents’ focus on task-relevant contexts by constructing a belief state via question answering. |
| Outcome: | The proposed method outperforms established baselines and achieves marked improvements on the BeIR zero-shot retrieval benchmark. |