Papers by Hanning Zhang
Mitigating the Alignment Tax of RLHF (2024.emnlp-main)
Copied to clipboard
Yong Lin, Hangyu Lin, Wei Xiong, Shizhe Diao, Jianmeng Liu, Jipeng Zhang, Rui Pan, Haoxiang Wang, Wenbin Hu, Hanning Zhang, Hanze Dong, Renjie Pi, Han Zhao, Nan Jiang, Heng Ji, Yuan Yao, Tong Zhang
| Challenge: | Large Language Models (LLMs) acquire a wide range of abilities during pre-training, but aligning LLMs under Reinforcement Learning with Human Feedback (RLHF) can lead to forgetting pretrained abilities, which is also known as the alignment tax. |
| Approach: | They propose to use a model averaging technique to find the most powerful alignment-forging Pareto front among RLHF algorithms. |
| Outcome: | The proposed method achieves the strongest alignment-forging Pareto front among competing methods. |
OpenGenAlign: A Preference Dataset and Benchmark for Trustworthy Reward Modeling in Open-Ended, Long-Context Generation (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing reward models perform suboptimal on held-out benchmarks, resulting in poor quality outputs. |
| Approach: | They propose a framework and a high-quality dataset to evaluate reward models . they define four key metrics to assess generation quality and develop a pipeline to evaluate outputs . |
| Outcome: | The proposed framework and dataset improves hallucination-free, comprehensive, reliable, and efficient open-ended long-context generation. |
R-Tuning: Instructing Large Language Models to Say ‘I Don’t Know’ (2024.naacl-long)
Copied to clipboard
Hanning Zhang, Shizhe Diao, Yong Lin, Yi Fung, Qing Lian, Xingyao Wang, Yangyi Chen, Heng Ji, Tong Zhang
| Challenge: | Existing methods for instruction tuning force the model to complete a sentence no matter whether it knows the knowledge or not. |
| Approach: | They propose a new approach to tuning large language models to refrain from answering questions beyond its parametric knowledge by identifying the disparity in parametric and parametric information. |
| Outcome: | The proposed approach improves a model’s ability to answer known questions and refrain from answering unknown questions. |
ScaleBiO: Scalable Bilevel Optimization for LLM Data Reweighting (2025.acl-long)
Copied to clipboard
Rui Pan, Dylan Zhang, Hanning Zhang, Xingyuan Pan, Minrui Xu, Jipeng Zhang, Renjie Pi, Xiaoyu Wang, Tong Zhang
| Challenge: | Existing paradigms for bilevel optimization require second-order information, making it difficult to scale them up. |
| Approach: | They propose a scalable instantiation of a bilevel optimization paradigm for large-scale LLMs by using a memory-efficient training technique. |
| Outcome: | The proposed paradigm scales to 30B-sized LLMs on 8H100 GPUs. |