Papers by Kai Zou
Capability Salience Vector: Fine-grained Alignment of Loss and Capabilities for Downstream Task Scaling Law (2025.acl-long)
Copied to clipboard
Qiming Ge, Shuhao Xing, Songyang Gao, Yunhua Zhou, Yicheng Zou, Songyang Zhang, Zhi Chen, Hang Yan, Qi Zhang, Qipeng Guo, Kai Chen
| Challenge: | Large language models have demonstrated impressive performance across a wide range of tasks, but this achievement comes with the trade-off of significant computational demands. |
| Approach: | They propose a scaling law that decomposes the overall validation loss and assigns different importance weights to tokens to assess a specific meta-capability. |
| Outcome: | The proposed model can predict the loss trending of models across different levels of computation without a gap between validation loss and model's downstream capabilities. |
Annotations Are Not All You Need: A Cross-modal Knowledge Transfer Network for Unsupervised Temporal Sentence Grounding (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing work on temporal sentence grounding rely on expensive video-query paired annotations . despite this, there are no ground-truth annotations in the current work . |
| Approach: | They propose to use paired video-query and segment boundary annotations to generate temporal sentence grounding without training. |
| Outcome: | The proposed model outperforms existing unsupervised methods and beats supervised ones on two challenging datasets. |
Unleashing the Reasoning Potential of LLMs by Critique Fine-Tuning on One Problem (2025.emnlp-main)
Copied to clipboard
| Challenge: | Critique Fine-Tuning (CFT) is a promising paradigm for unlocking the reasoning capabilities of large language models. |
| Approach: | They propose a method that leverages critique data generated from a single math problem to improve reasoning accuracy. |
| Outcome: | The proposed method surpasses one-shot RLVR while requiring 15 to 20 times less compute. |
SWE-Fixer: Training Open-Source LLMs for Effective and Efficient GitHub Issue Resolution (2025.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models excel in code generation benchmarks, but these benchmarks focus on single-file scenarios with constrained context scope. |
| Approach: | They propose an open-source framework to effectively resolve GitHub issues using a code file retrieval module and a model-based code editing module. |
| Outcome: | The proposed approach achieves state-of-the-art performance on two GitHub benchmarks. |
EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks (D19-1)
Copied to clipboard
| Challenge: | Existing data augmentation techniques for text classification are difficult to implement and cost a high amount of money. |
| Approach: | They propose to use four simple but powerful operations to boost performance on text classification tasks to improve synonym replacement, random insertion, random swap, and random deletion. |
| Outcome: | The proposed techniques improve performance on five classification tasks and are particularly useful for smaller datasets. |
What are the Essential Factors in Crafting Effective Long Context Multi-Hop Instruction Datasets? Insights and Best Practices (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods to generate long-context instruction-tuning data are limited by poor quality and fewer than 35% of samples are multi-hop . |
| Approach: | They propose a framework that integrates a quality verification agent, a single-hop question generation agent, and a multi-hop questions merger agent to enhance model performance. |
| Outcome: | The proposed framework significantly improves data quality with high-quality, multi-hop, and diverse data. |
Uni-MMMU: A Massive Multi-discipline Multimodal Unified Benchmark (2026.acl-long)
Copied to clipboard
Kai Zou, Ziqi Huang, Yuhao Dong, Shulin Tian, Dian Zheng, Hongbo Liu, Jingwen He, Bin Liu, Yu Qiao, Ziwei Liu
| Challenge: | Existing evaluations treat visual understanding and generation in isolation or overlook tasks that inherently couple them. |
| Approach: | They propose a benchmark that examines the bidirectional synergy between generation and understanding across eight reasoning-centric domains. |
| Outcome: | The proposed model systematically unfolds the bidirectional synergy between generation and understanding across eight reasoning-centric domains. |
VisualWebInstruct: Scaling up Multimodal Instruction Data through Web Search (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing vision-language models struggle with reasoning-focused tasks due to the lack of high-quality training data. |
| Approach: | They propose a new approach that leverages search engines to create a multimodal multimodal dataset . they use a set of 30,000 seed images to extract HTML data from 700K unique URLs . |
| Outcome: | The proposed model achieves the best known performance on MMMU-Pro (40.7), MathVerse (42.6), and DynaMath (55.7). |
VisCoder: Fine-Tuning LLMs for Executable Python Visualization Code Generation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing instruction-tuning datasets lack execution-grounded supervision and offer limited support for iterative code correction. |
| Approach: | They propose a large-scale instruction tuning dataset for Python-based visualization and self-correction. |
| Outcome: | The proposed dataset outperforms strong open-source baselines and proprietary models like GPT-4o-mini. |
BrowseComp-Plus: A Fair and Disentangled Evaluation Benchmark for Deep Search Agents (2026.acl-long)
Copied to clipboard
Zijian Chen, Xueguang Ma, Shengyao Zhuang, Ping Nie, Kai Zou, Sahel Sharifymoghaddam, Andrew Liu, Joshua Green, Kshama Patel, Ruoxi Meng, Mingyi Su, Yanxi Li, Haoran Hong, Xinyu Shi, Xuye Liu, Hosna Oyarhoseini, Nandan Thakur, Crystina Zhang, Luyu Gao, Wenhu Chen, Jimmy Lin
| Challenge: | Existing benchmarks for deep search agents rely on blackbox web search APIs . dynamic and opaque web APIs hinder reproducibility and fair comparisons - authors . |
| Approach: | They propose a benchmark that employs a fixed corpus for controlled retrieval for deep search agents. |
| Outcome: | The new benchmark shows that agents that combine large language models with retrieval tools excel at complex, reasoning-intensive queries. |
D4: a Chinese Dialogue Dataset for Depression-Diagnosis-Oriented Chat (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing human-machine dialogue systems are not able to provide diagnostic information for depression diagnosis due to stigma associated with mental illness. |
| Approach: | They propose to construct a Chinese Dialogue Dataset for depression-diagnosis-oriented chat based on clinical depression diagnostic criteria. |
| Outcome: | The proposed system can be used to diagnose depression using a Chinese Dialogue Dataset. |
Evaluation Metrics in the Era of GPT-4: Reliably Evaluating Large Language Models on Sequence to Sequence Tasks (2023.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) evaluation is a patchy and inconsistent landscape . established automatic evaluation metrics are poor surrogates, correlating weakly with human judgement. |
| Approach: | They propose to use both automatic and human evaluation to evaluate generative LLMs on three NLP benchmarks: text summarisation, text simplification and grammatical error correction. |
| Outcome: | The proposed model outperforms many popular models according to human reviewers on the majority of metrics, while scoring much worse when using classic automatic evaluation metrics. |