Papers by Jiaxi Li
CodeArena: Evaluating and Aligning CodeLLMs on Human Preference (2025.emnlp-main)
Copied to clipboard
Jian Yang, Jiaxi Yang, Wei Zhang, Jin Ke, Yibo Miao, Lei Zhang, Liqun Yang, Zeyu Cui, Yichang Zhang, Zhoujun Li, Binyuan Hui, Junyang Lin
| 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. |
HELENE: Hessian Layer-wise Clipping and Gradient Annealing for Accelerating Fine-tuning LLM with Zeroth-order Optimization (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) face memory challenges due to the high cost of backpropagation. |
| Approach: | They propose a zeroth-order (ZO) optimization that matches memory usage to inference . they propose scalable and memory-efficient zeroth order (ZE) optimizer that integrates annealed A-GNB gradients with diagonal Hessian estimation and layer-wise clipping as a second-order pre-conditioner. |
| Outcome: | The proposed algorithm outperforms state-of-the-art methods with an average speedup of 20 over MeZO on RoBERTa-large and OPT-1.3B. |
SciAssess: Benchmarking LLM Proficiency in Scientific Literature Analysis (2025.findings-naacl)
Copied to clipboard
Hengxing Cai, Xiaochen Cai, Junhan Chang, Sihang Li, Lin Yao, Wang Changxin, Zhifeng Gao, Hongshuai Wang, Li Yongge, Mujie Lin, Shuwen Yang, Jiankun Wang, Mingjun Xu, Jin Huang, Xi Fang, Jiaxi Zhuang, Yuqi Yin, Yaqi Li, Changhong Chen, Zheng Cheng, Zifeng Zhao, Linfeng Zhang, Guolin Ke
| Challenge: | Existing benchmarks fail to adequately evaluate the proficiency of Large Language Models (LLMs) Existing standards do not cover the skills needed to evaluate LLMs in scientific literature analysis. |
| Approach: | They propose a benchmark to evaluate the proficiency of large language models in scientific literature analysis. |
| Outcome: | SciAssess evaluates 11 LLMs on multiple tasks across scientific fields. |
Through the Valley: Path to Effective Long CoT Training for Small Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Long chain-of-thought (CoT) supervision is effective for large language models . but small models trained on limited long CoT data experience performance degradation . |
| Approach: | They identify a phenomenon called Long CoT Degradation in small language models . long CoT data can be used to generate long chain-of-thought (CoT) responses . |
| Outcome: | The results show that models trained on 8k long CoT examples lose up to 75% of their original performance before fine-tuning. |
Video-LLaVA: Learning United Visual Representation by Alignment Before Projection (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to visual-language understanding lack unified tokenization for images and videos . lack of unified visual representations makes it difficult to learn multi-modal interactions from poor projection layers. |
| Approach: | They propose to unify visual representation into the language feature space to advance the foundational LLM towards a unified LVLM. |
| Outcome: | The proposed model outperforms Video-ChatGPT on image benchmarks and on 9 image benchmark benchmarks. |
Contextual Distortion Reveals Constituency: Masked Language Models are Implicit Parsers (2023.acl-long)
Copied to clipboard
| Challenge: | a novel chart-based method for extracting parse trees from masked language models is proposed . a graph-based approach can be used to extract parser trees without training separate parsers . |
| Approach: | They propose a chart-based method for extracting parse trees from masked language models . they use a set of perturbations motivated by the linguistic concept of constituency tests to score each span . |
| Outcome: | The proposed method outperforms state-of-the-art methods on english with masked LMs and in multilingual settings. |
START: Self-taught Reasoner with Tools (2025.emnlp-main)
Copied to clipboard
Chengpeng Li, Mingfeng Xue, Zhenru Zhang, Jiaxi Yang, Beichen Zhang, Bowen Yu, Binyuan Hui, Junyang Lin, Xiang Wang, Dayiheng Liu
| 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%). |
One-Shot Learning as Instruction Data Prospector for Large Language Models (2024.acl-long)
Copied to clipboard
Yunshui Li, Binyuan Hui, Xiaobo Xia, Jiaxi Yang, Min Yang, Lei Zhang, Shuzheng Si, Ling-Hao Chen, Junhao Liu, Tongliang Liu, Fei Huang, Yongbin Li
| Challenge: | Contemporary practices in instruction tuning often hinge on enlarging data scaling without a clear strategy for ensuring data quality. |
| Approach: | They propose a method that leverages one-shot learning to discern and select high-quality instruction data from extensive datasets. |
| Outcome: | Nuggets outperforms existing methods on MT-Bench and Alpaca-Eval benchmarks. |
Mnemis: Dual-Route Retrieval on Hierarchical Graphs for Long-Term LLM Memory (2026.acl-long)
Copied to clipboard
Zihao Tang, Xin Yu, Ziyu Xiao, Zengxuan Wen, Zelin Li, Jiaxi Zhou, Hualei Wang, Haohua Wang, Haizhen Huang, Weiwei Deng, Feng Sun, Qi Zhang
| Challenge: | Existing methods for retrieving historical messages are based on similarity-based mechanisms. |
| Approach: | They propose a system that integrates System-1 similarity search with a complementary System-2 mechanism, termed Global Selection. |
| Outcome: | The proposed framework achieves state-of-the-art on long-term memory benchmarks and 93.9 on LoCoMo and 91.6 on LongMemEval-S. |
Unraveling Feature Extraction Mechanisms in Neural Networks (2023.emnlp-main)
Copied to clipboard
| Challenge: | Neural networks have become indispensable across a variety of natural language processing tasks. |
| Approach: | They propose a theoretical approach based on Neural Tangent Kernels to investigate neural networks' internal mechanisms. |
| Outcome: | The proposed approach can be applied to analyze language modeling tasks . it shows that the choice of activation function can affect feature extraction . |
Marathon: A Race Through the Realm of Long Context with Large Language Models (2024.acl-long)
Copied to clipboard
| Challenge: | Existing long-context benchmarks do not accurately evaluate large language models’ comprehension and reasoning abilities in extended texts. |
| Approach: | They propose a new evaluation benchmark that adopts a multiple-choice question format and uses a multi-choke question format to assess the comprehension and reasoning skills of large language models. |
| Outcome: | The proposed benchmark provides a rapid, precise, and unbiased appraisal of the long-context comprehension skills of large language models. |
KiPT: Knowledge-injected Prompt Tuning for Event Detection (2022.coling-1)
Copied to clipboard
| Challenge: | Existing prompt-based methods may suffer from low precision because they lack event-related semantic knowledge. |
| Approach: | They propose a Knowledge-injected Prompt Tuning model to improve prompt tuning . event detection aims to detect events from text by identifying and classifying event triggers . |
| Outcome: | The proposed model outperforms baseline models in few-shot scenarios. |
Decomposed Prompt Tuning via Low-Rank Reparameterization (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Pre-trained language models have achieved remarkable performance on various tasks. |
| Approach: | They propose a decomposed prompt tuning approach that utilizes low-rank matrices to initialize the soft prompt. |
| Outcome: | The proposed method significantly reduces the number of trainable parameters while maintaining effectiveness. |
LM-Interview: An Easy-to-use Smart Interviewer System via Knowledge-guided Language Model Exploitation (2024.emnlp-demo)
Copied to clipboard
| Challenge: | Semi-structured interviews are a crucial method of data acquisition in qualitative research. |
| Approach: | They propose a semi-structured interview system that automates interview preparation, analysis and control by interviewers. |
| Outcome: | Experimental results show that LM-Interview performs comparable to human interviewers . the system can be used to analyze semi-structured interviews without interviewers' involvement . |
Relaxing the Constraints: A Dual-Importance Projection Mechanism for Lifelong Model Editing (2026.findings-acl)
Copied to clipboard
Zhenghai Chen, Senbin Xu, Jiaxi Tan, Xinhua Wu, Yan Zhang, Xiawu Zheng, Shengchuan Zhang, Ke Li, Sicheng Zhao, Liujuan Cao, Rongrong Ji
| Challenge: | Existing knowledge editing methods rely on strict orthogonal projection to preserve previously edited knowledge, but this constraint limits gradient expressiveness, resulting in degradation of model generalization and overall performance as the number of edits increases. |
| Approach: | They propose a method that leverages Singular Value Decomposition to identify critical gradient subspaces and introduces a dual mechanism comprising "accumulated importance" and "projection importance" |
| Outcome: | Extensive experiments on five mainstream LLMs show that the proposed method achieves an average comprehensive performance improvement of 10.36% and effectively maintains the model’s general capabilities on downstream tasks. |