Zhepei Hong

2026

July

1. arXiv

   

   EasyOPD: An Easy-to-use On-Policy Distillation Framework for Large Language Models

   Jie Sun, Mao Zheng, Mingyang Song, Qiyong Zhong, Gengsheng Li, Zhepei Hong, Chang Wu, Pengfei Liu, Junfeng Fang, and Xiang Wang

   arXiv preprint, Jul 2026

   Preprint, Co-author

   Abs arXiv PDF Code

   Conventional language-model distillation often relies on fixed teacher-generated data, which may not cover the states encountered by an evolving student policy. On-policy distillation (OPD) instead collects teacher or evaluator supervision on student-generated rollouts. However, existing OPD methods differ substantially in supervision form, tokenizer compatibility, teacher access, and supervision granularity, leading to fragmented implementations that are difficult to reproduce and extend. We present EasyOPD, an on-policy distillation framework built on verl, a distributed reinforcement-learning framework for large language models. EasyOPD separates user-side configuration, method-specific supervision logic, and verl-based execution. Its method modules connect to the shared backend through extension boundaries for loss construction, rollout metadata, reward processing, tokenizer alignment, and teacher-side computation. We instantiate representative methods for three OPD settings – cross-tokenizer OPD, on-policy self-distillation, and step-wise OPD. Experiments on reasoning, code-generation, scientific-knowledge, and tool-use benchmarks show that these implementations can be executed through the same verl-based backend while retaining their method-specific objectives and task-dependent performance profiles.

May

1. arXiv

   

   TRACE: Trajectory Risk-Aware Compression for Long-Horizon Agent Safety

   Zhepei Hong, Lin Wang, Liting Li, Haokai Ma, Junfeng Fang, Fei Shen, Dan Zhang, and Xiang Wang

   arXiv preprint, May 2026

   Preprint, First author

   Abs arXiv PDF Code

   Long-horizon LLM agents produce safety evidence across long trajectories, where sparse, delayed, and compositional risk signals often escape local moderation. We reframe long-horizon agent safety detection as trajectory-level evidence compression and propose Trajectory Risk-Aware Compression for Long-Horizon Agent Safety (TRACE). TRACE uses a Compressor-Reader design: the Compressor encodes the full trajectory into a compact latent evidence state under trajectory-level supervision, and the Reader judges the raw trajectory with this latent evidence state as a safety reference. Across ASSE-Bench, Pre-Ex-Bench, and R-Judge, TRACE achieves the best accuracy on all evaluated backbones, improving over strong baselines by up to 12.6 percentage points. On LongSafety, TRACE shows smaller performance degradation as context length grows.

2. arXiv

   

   Rubric-based On-policy Distillation

   Junfeng Fang^*, Zhepei Hong^*, Mao Zheng, Mingyang Song, Gengsheng Li, Houcheng Jiang, Dan Zhang, Haiyun Guo, Xiang Wang, and Tat-Seng Chua

   arXiv preprint, May 2026

   Preprint, * Co-first author

   Abs arXiv PDF Code

   On-policy distillation (OPD) is a powerful paradigm for model alignment, yet its reliance on teacher logits restricts its application to white-box scenarios. We introduce ROPD, a rubric-based OPD framework that induces prompt-specific rubrics from teacher-student contrasts and uses them to score student rollouts for on-policy optimization. Empirically, ROPD outperforms advanced logit-based OPD methods across most scenarios and achieves up to a 10x gain in sample efficiency, positioning rubric-based OPD as a flexible, black-box-compatible alternative for scalable distillation across proprietary and open-source LLMs.

April

1. BSPC

   

   HEAT: Hierarchical Emotion Adaptation with Progressive Thresholding for EEG Emotion and Consciousness Detection

   Zhepei Hong, Rongtao Chen, Liting Li, Jiajun Chen, Wei Gao, and Jiahui Pan

   Biomedical Signal Processing and Control, Apr 2026

   Accepted, First author

   PDF Code

March

1. TAFFC

   

   Multi-scale Dynamic Temporal Network with Graph Matching Domain Adaptation for Cross-Subject EEG Emotion Recognition

   Rongtao Chen^*, Zhepei Hong^*, Liting Li, Zhuoyi Huang, Jingcong Li, and Jiahui Pan

   IEEE Transactions on Affective Computing, Mar 2026

   Accepted, * Co-first author

   PDF Code

2025

October

1. BIBM

   

   PR-DA: Prototype Regularization Domain Adaptation for Cross-Subject EEG-Based Emotion Recognition

   Rongtao Chen^*, Zhepei Hong^*, Qi You, Chuwen Xie, Jiahui Zhang, and Jiahui Pan

   In IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Oct 2025

   Accepted, Oral presentation, * Co-first author

   PDF Code