Author

Peng-Fei Wan

h-index 11 View on Semantic Scholar

We have 6 of 36 papers

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Preprint Oct 2026

RVG-Bench: A Benchmark for Reasoning-Driven Audio-Visual Temporal Grounding

Existing temporal video grounding (TVG) benchmarks primarily focus on localizing target video segments according to descriptive text queries of the visual content. However, real-world queries are more diverse and often require the model to first understand the user's intent before localization, such as finding visual e...

Yue Feng, Bo-Shen Xu, Zihao Yue et al. · 0 citations

#artificial intelligence Preprint Oct 2026

EventLens: Event-Structure Reinforcement Learning for Video Understanding

Recent multimodal large language models (MLLMs) achieve strong results on video question answering, yet often fail to recover the temporal structure that makes a video coherent. They may merge neighboring events, miss single-event progression, or infer implausible relations across events. We study structured temporal u...

Guo-Wen Zhang, Bo-Shen Xu, Zihao Yue et al. · 0 citations

#machine learning Preprint Oct 2026

ORCA: The Annealed Spectral Conditioning Optimizer for Faster, Better LLM Training

This work introduces ORCA (Orthogonal Regularization, Cooled After), a minimal optimizer intervention that applies strong but temporary soft orthogonality regularization early in training, then removes it, which allows the weights to benefit from a broader spectrum early on and adapt freely afterward.

Yuan-Shi Liu, Bo-Yuan Jiang, Liang Hou et al. · 0 citations

Preprint Aug 2026

Manifold Drift in Flow Preference Optimization: A Root Cause of Reward Hacking

Preference optimization is a standard alignment method for generative models, yet extending it to continuous-time dynamics remains non-trivial. In flow matching, reward-driven updates modify transport trajectories without an inherent constraint to the pretrained data manifold and can move terminal samples off the pretr...

Yan-Sen Han, Shengyi Liao, Yuan-Xing Zhang et al. · 0 citations

Preprint Aug 2026

PlayWorld: Benchmarking World Models with Agent Players over Long-Horizon Objectives

PlayWorld is introduced, a benchmark providing 171 scenarios, each with a specified objective, to assess models along four core dimensions: geometry consistency, interaction fidelity, out-of-sight evolution, and insight evolution, and basic ability metrics for video quality and controllability.

Kai Ding, Xi Chen, Ming-Hong Cai et al. · 3 citations

#artificial intelligence Preprint Sep 2026

FlowCPO: A Unified Divergence View of Preference Alignment for Flow Models

This work introduces FlowCPO, an offline forward-KL objective that uses both preferred and dispreferred samples without online rollouts and shows under explicit regularity conditions that the forward-KL objective is bounded by a contrastive flow matching loss, yielding a tractable surrogate on fixed data.

Yan-Sen Han, Shengyi Liao, Peng Sun et al. · 1 citation