Papers by Junyang Zhang

28 papers
CodeArena: Evaluating and Aligning CodeLLMs on Human Preference (2025.emnlp-main)

Copied to clipboard

Challenge: Code large language models (codeLLMs) focus on synthesizing the correct code snippet, ignoring the alignment with human preferences.
Approach: They propose a benchmark code-based on 40 categories and 44 programming languages to emulate real-world coding tasks.
Outcome: The proposed benchmarks show that open-source code LLMs perform better than open-sourced ones.
Turning the Tide: Repository-based Code Reflection (2025.findings-emnlp)

Copied to clipboard

Challenge: Code large language models (LLMs) enhance programming by understanding and generating code across languages.
Approach: a new benchmark evaluates code understanding and generation in repositories using code large language models.
Outcome: The proposed model improves code understanding and generation in repositories by evaluating 1,888 test cases across 6 programming languages.
FPE2M2: Approaching Lossless and Efficient Quantization with Native Floating Point (2025.findings-acl)

Copied to clipboard

Challenge: Auto-regressive decoding is a memory-bound job, meaning decoding performance is limited by the bandwidth rather than the computational capabilities of the GPU.
Approach: They propose a framework that supports lossless weight-only quantization inference and validate it on Qwen and LLaMA Models.
Outcome: The proposed framework achieves the highest efficiency with lossless accuracy on Qwen and LLaMA Models across various modalities.
NOVA-63: Native Omni-lingual Versatile Assessments of 63 Disciplines (2025.emnlp-main)

Copied to clipboard

Challenge: Existing multilingual benchmarks show severe drawbacks, such as overly translated content, the absence of difficulty control, and disciplinary imbalance, making the benchmarking process unreliable and showing low convincingness.
Approach: They propose a multilingual benchmark that integrates LLM-assisted formatting, expert quality verification, and multi-level difficulty screening to provide a comprehensive, difficult multilingual assessment.
Outcome: The proposed benchmark features 93,536 questions sourced from native speakers across 14 languages and 63 academic disciplines.
Can Large Language Models Always Solve Easy Problems if They Can Solve Harder Ones? (2024.emnlp-main)

Copied to clipboard

Challenge: Large language models (LLMs) have impressive capabilities, but still suffer from inconsistency issues.
Approach: They develop a ConsisEval benchmark to evaluate LLMs' inconsistency . they find that LLM models can paradoxically fail at easier problems .
Outcome: The proposed model achieves highest consistency score but inconsistent to specific questions due to distraction by redundant information, misinterpretation of questions, etc.
ProcessBench: Identifying Process Errors in Mathematical Reasoning (2025.acl-long)

Copied to clipboard

Challenge: Existing models fail to generalize to more challenging math problems, authors say . existing benchmarks related to assessing language models' reasoning process are limited .
Approach: They propose a tool to measure language models' ability to identify erroneous steps in reasoning . they use two types of models: process reward models and critic models .
Outcome: The proposed model outperforms existing models in evaluating language models' reasoning process . the best open-source model has demonstrated the critique capability competitive with the proprietary model .
RMTBench: Benchmarking LLMs Through Multi-Turn User-Centric Role-Playing (2025.findings-emnlp)

Copied to clipboard

Challenge: Existing benchmarks focus on character-centric approach and fail to reflect real-world applications.
Approach: RMTBench is a user-centric bilingual role-playing benchmark featuring 80 diverse characters and over 8,000 dialogue rounds.
Outcome: RMTBench features 80 diverse characters and over 8,000 dialogue rounds.
Confidence v.s. Critique: A Decomposition of Self-Correction Capability for LLMs (2025.acl-long)

Copied to clipboard

Challenge: Existing approaches to improve self-correction performance of Large Language Models are based on intrinsic selfcorrectione, which allows the model to check and revise its selfgenerated answers without external feedback.
Approach: They propose to decompose the self-correction capability into confidence and critique capabilities and a metric for overall self-corretion capability evaluation.
Outcome: The proposed method outperforms vanilla SFT and achieves much higher accuracy after self-correction.
Outcome Accuracy is Not Enough: Aligning the Reasoning Process of Reward Models (2026.acl-long)

Copied to clipboard

Challenge: Recent studies observe a phenomenon where reward models achieve high accuracy on static datasets but fail to generalize effectively during RLHF.
Approach: They propose a method that combines rationale consistency with outcome accuracy to improve performance on RM-Bench and JudgeBench.
Outcome: The proposed method surpasses baselines on RM-Bench and JudgeBench by an average of 5% and improves creative writing tasks by 7%.
PLAWBENCH: A Rubric-Based Benchmark for Evaluating LLMs in Real-World Legal Practice (2026.acl-long)

Copied to clipboard

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.
START: Self-taught Reasoner with Tools (2025.emnlp-main)

Copied to clipboard

Challenge: Large Reasoning Models (LRMs) have demonstrated remarkable capabilities in complex reasoning through long chain-of-thought, yet they struggle with precise computations and algorithmic operations.
Approach: They propose a training-free approach that activates LRMs’ latent tool-use capabilities through artificial hints and a framework that enables models to learn effective tool utilization through diverse hint patterns and rejection-based data synthesis.
Outcome: Experiments show that START significantly improves state-of-the-art LRMs across challenging benchmarks, including competition-level mathematics (AMC23: 95.0%, AIME24: 75.6%) and graduate-level science questions (GPQA: 64.6%).
ReSel: N-ary Relation Extraction from Scientific Text and Tables by Learning to Retrieve and Select (2022.emnlp-main)

Copied to clipboard

Challenge: Our proposed method extracts N-ary relation tuples from scientific articles.
Approach: They propose a method that decomposes the task into two stages . they propose modal query and modal entity selection . their results show that ReSel outperforms state-of-the-art baselines significantly .
Outcome: The proposed method outperforms state-of-the-art baselines on three scientific information extraction datasets.
P-MMEval: A Parallel Multilingual Multitask Benchmark for Consistent Evaluation of LLMs (2025.emnlp-main)

Copied to clipboard

Challenge: Recent advances in large language models showcase varied multilingual capabilities across tasks . previous assessments focused on fundamental natural language processing (NLP) or isolated capability-specific tasks.
Approach: They propose a multilingual multitask benchmark to assess multilingual capabilities . they use a large-scale benchmark covering fundamental and capability-specialized datasets .
Outcome: The proposed benchmark compares models and tasks across languages and tasks and examines knowledge transfer from English to other languages.
The Lessons of Developing Process Reward Models in Mathematical Reasoning (2025.findings-acl)

Copied to clipboard

Challenge: a recent study shows that process reward models can make mistakes, leading to wrong conclusions.
Approach: They propose a consensus filtering mechanism that integrates MC estimation with LLM-as-a-judge to improve model performance and data efficiency.
Outcome: The proposed model outperforms existing open-source alternatives and provides practical guidelines for future research.
Sketch and Refine: Towards Faithful and Informative Table-to-Text Generation (2021.findings-acl)

Copied to clipboard

Challenge: Existing methods for table-to-text generation suffer from poor faithfulness and low coverage.
Approach: They propose a method that combines Autoregressive and Non-Autoregressive generation to generate a table-to-text from a key-value table using a skeleton and an edit-based non-autoregressively generation model.
Outcome: The proposed method outperforms the existing methods on WikiPerson and WikiBio datasets on coverage and faithfulness.
UNIKIE-BENCH: Benchmarking Large Multimodal Models for Key Information Extraction in Visual Documents (2026.acl-long)

Copied to clipboard

Challenge: Recent Large Multimodal Models (LMMs) have shown promising potential for performing end-to-end KIE directly from document images.
Approach: They propose a benchmark to evaluate the performance of Large Multimodal Models (LMMs) using a constrained-category KIE track and an open-categorical KIE Track.
Outcome: Experiments on 15 state-of-the-art LMMs show performance degradation under diverse schema definitions, long-tail key fields, and complex layouts, along with pronounced performance disparities across different document types and scenarios.
HellaSwag-Pro: A Large-Scale Bilingual Benchmark for Evaluating the Robustness of LLMs in Commonsense Reasoning (2025.findings-acl)

Copied to clipboard

Challenge: Existing studies show that large language models are robust in commonsense reasoning . however, some variations in questions can lead to incorrect responses .
Approach: They propose a large-scale bilingual benchmark consisting of 11,200 cases . they conduct extensive experiments on 41 representative LLMs .
Outcome: The proposed benchmark systematically evaluates the robustness of large language models in commonsense reasoning.
Disentangling Reasoning Tokens and Boilerplate Tokens For Language Model Fine-tuning (2025.findings-acl)

Copied to clipboard

Challenge: Existing approaches to enhance agent capabilities for Large Language Models treat all tokens equally . however, reasoning tokens versus boilerplate tokens differ in importance and learning complexity . recent research has focused on enhancing agent capabilities in large language models .
Approach: They propose a Shuffle-Aware Discriminator (SHAD) for adaptive token discrimination . they propose SHAD method which adaptively emphasizes reasoning tokens during fine-tuning .
Outcome: The proposed method improves performance over standard fine-tuning methods.
A Probabilistic Inference Scaling Theory for LLM Self-Correction (2025.emnlp-main)

Copied to clipboard

Challenge: Large Language Models (LLMs) have demonstrated the capability to refine their generated answers through self-correction, enabling continuous performance improvement over multiple rounds.
Approach: They propose a probabilistic theory to model the dynamics of accuracy change and explain performance improvements observed in multi-round self-correction.
Outcome: The proposed model can predict accuracy curves and improve accuracy over multiple rounds.
Qwen2.5-xCoder: Multi-Agent Collaboration for Multilingual Code Instruction Tuning (2025.acl-long)

Copied to clipboard

Challenge: Existing methods to train code LLMs view each programming language in isolation . experimental results show that Qwen2.5-xCoder can bridge the gap between different programming languages .
Approach: They propose a framework that allows agents to collaborate to enhance multilingual instruction tuning for code LLMs.
Outcome: Experimental results show that Qwen2.5-xCoder can transfer knowledge efficiently and effectively between languages.
DeepPlanning: Benchmarking Long-Horizon Agentic Planning with Verifiable Constraints (2026.acl-long)

Copied to clipboard

Challenge: Existing LLM planning benchmarks emphasize local, step-level reasoning rather than global constrained optimization.
Approach: They propose a benchmark for practical long-horizon agent planning that uses local constrained reasoning and global constrained optimization.
Outcome: The proposed benchmarks show that even frontier agentic LLMs struggle with these problems.
Towards Knowledge-Based Recommender Dialog System (D19-1)

Copied to clipboard

Challenge: Existing frameworks that only provide information about user preferences can be inaccurate in e-commerce recommender systems.
Approach: They propose a framework which integrates the recommender system and dialog generation system by introducing information about users’ preferences.
Outcome: The proposed framework can achieve better performance in both dialog generation and recommendation compared with baselines.
RemoteRAG: A Privacy-Preserving LLM Cloud RAG Service (2025.findings-acl)

Copied to clipboard

Challenge: Large language models (LLMs) have a tendency to generate factually incorrect or purely fictional responses, a phenomenon known as hallucination.
Approach: They propose to use remote RAG to protect user query from privacy leakage . they introduce (n,)-DistanceDP to characterize privacy leakages of user query .
Outcome: The proposed solution can resist embedding inversion attacks while achieving no loss in retrieval under various settings.
HSS-Synth: Humanities and Social Sciences Data Synthesis for LLMs (2026.findings-acl)

Copied to clipboard

Challenge: High-quality, diverse data are vital for large language models (LLMs) but remain scarce and costly.
Approach: They define the first HSS domain system covering 14 mainstream fields and introduce HSS-Synth.
Outcome: the proposed pipeline outperforms 14 leading baselines on 16 benchmarks.
From Completion to Editing: Unlocking Context-Aware Code Infilling via Search-and-Replace Instruction Tuning (2026.acl-long)

Copied to clipboard

Challenge: Fill-in-the-Middle (FIM) models suffer from performance degradation and prohibitive latency.
Approach: They propose a search-and-replace infilling framework that integrates agentic verification and editing into a single-pass inference process.
Outcome: The proposed framework harmonizes completion tasks with the instruction-following priors of Chat LLMs, extending the paradigm from static infilling to dynamic context-aware editing.
Transferring General Multimodal Pretrained Models to Text Recognition (2023.findings-acl)

Copied to clipboard

Challenge: Existing methods for text recognition rely on large-scale pretraining on human-annotated or synthetic data.
Approach: They propose a method to transfer multimodal pretrained models to text recognition using image captioning.
Outcome: The proposed method outperforms the baselines and achieves state-of-the-art performance in the Chinese text recognition benchmark.
SpeechLLM-as-Judges: Towards General and Interpretable Speech Quality Evaluation (2026.acl-long)

Copied to clipboard

Challenge: Existing methods for evaluating the perceptual quality of synthetic speech are limited due to the complexity of perceptual quality factors and the diversity of speech generation tasks.
Approach: They propose a new paradigm for enabling large language models to conduct structured speech quality evaluation using a large-scale dataset.
Outcome: The proposed model performs well across tasks and languages.
LightMoE: Task-Aware Expert Availability Management for Memory-Efficient MoE-LLM Inference (2026.findings-acl)

Copied to clipboard

Challenge: Existing solutions for balancing model accuracy with inference latency are limited due to memory constraints.
Approach: They propose a framework for memory-efficient MoE inference that exploits the functional redundancy and temporal locality of expert activation.
Outcome: The proposed framework improves accuracy-efficiency trade-off by 4.3% over pruned models and 2.4% over dynamic swapping methods while maintaining inference latency comparable to pruned model.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations