11 papers
Can VLMs Truly Forget? Benchmarking Training-Free Visual Concept Unlearning
Zhangyun Tan, Zeliang Zhang, Susan Liang +3
VLMs trained on web-scale data retain sensitive and copyrighted visual concepts that deployment may require removing. Training-based unlearning methods share a structural flaw: fin…
Why Instruction-Based Unlearning Fails in Diffusion Models?
Zeliang Zhang, Rui Sun, Jiani Liu +2
Instruction-based unlearning has proven effective for modifying the behavior of large language models at inference time, but whether this paradigm extends to other generative model…
Training Large Reasoning Models Efficiently via Progressive Thought Encoding
Zeliang Zhang, Xiaodong Liu, Hao Cheng +3
Large reasoning models (LRMs) excel on complex problems but face a critical barrier to efficiency: reinforcement learning (RL) training requires long rollouts for outcome-based rew…
VidComposition: Can MLLMs Analyze Compositions in Compiled Videos?
Yolo Y. Tang, Junjia Guo, Hang Hua +9
The advancement of Multimodal Large Language Models (MLLMs) has enabled significant progress in multimodal understanding, expanding their capacity to analyze video content. However…
AdvEvo-MARL: Shaping Internalized Safety through Adversarial Co-Evolution in Multi-Agent Reinforcement Learning
Zhenyu Pan, Yiting Zhang, Zhuo Liu +13
LLM-based multi-agent systems excel at planning, tool use, and role coordination, but their openness and interaction complexity also expose them to jailbreak, prompt-injection, and…
Diversifying the Expert Knowledge for Task-Agnostic Pruning in Sparse Mixture-of-Experts
Zeliang Zhang, Xiaodong Liu, Hao Cheng +2
By increasing model parameters but activating them sparsely when performing a task, the use of Mixture-of-Experts (MoE) architecture significantly improves the performance of Large…