Pengkun Jiao「焦鹏昆」
Ph.D. Student · Computer Science · Fudan University
I am a Ph.D. student at the School of Computer Science, Fudan University, advised by Prof. Jingjing Chen and Prof. Yu-Gang Jiang. I also work under the guidance of Prof. Na Zhao and Prof. Bin Zhu.
My research focuses on multimodal learning, large language models, reinforcement learning, and efficient AI. I study reliable multimodal models and efficient adaptation and inference.
News
| Jul 11, 2026 | One paper addressing structural bias in attention mechanisms is accepted to ACM MM 2026. |
|---|---|
| May 01, 2026 | One paper about KV Cache Eviction is accepted to ICML 2026. |
| Apr 16, 2026 | One paper on LoRA-MoE is accepted to ICMR 2026. |
| Apr 06, 2026 | One paper on induced hallucinations in Video-LLMs is accepted to the Findings of ACL 2026. |
| Jun 26, 2025 | One paper about multi-modal large language models is accepted to ICCV 2025. |
Selected Publications
- Disentangling Semantic Attention from Structural Bias in the Attention ManifoldACM MM, 2026
- Predicting Future Utility: Global Combinatorial Optimization for Task-Agnostic KV Cache EvictionICML, 2026
- Spatiotemporal Sycophancy: Negation-Based Gaslighting in Video Large Language ModelsACL Findings, 2026
-
Don’t Deceive Me: Mitigating Gaslighting through Attention Reallocation in LMMshttps://arxiv.org/abs/2504.09456, 2025 -
From Holistic to Localized: Local Enhanced Adapters for Efficient Visual Instruction Fine-TuningICCV, 2025 -
-
Unlocking Textual and Visual Wisdom: Open-Vocabulary 3D Object Detection Enhanced by Comprehensive Guidance from Text and ImageECCV, 2024 -
Domain Expansion and Boundary Growth for Open-Set Single-Source Domain GeneralizationTMM, 2024
