10 papers
More Is Not Always Better: Cross-Component Interference in LLM Agent Scaffolding
Ming Liu
LLM agent systems are built by stacking scaffolding components (planning, tools, memory, self-reflection, retrieval) assuming more is better. We study cross-component interference…
Decodable but Not Corrected by Fixed Residual-Stream Linear Steering: Evidence from Medical LLM Failure Regimes
Ming Liu
Can linearly decodable failure signals in LLM hidden states be leveraged to correct those failures? We investigate this classification-correction gap via Overthinking (OT)--a stabl…
Wan-R1: Verifiable-Reinforcement Learning for Video Reasoning
Ming Liu, Yunbei Zhang, Shilong Liu +2
Video generation models produce visually coherent content but struggle with tasks requiring spatial reasoning and multi-step planning. Reinforcement learning (RL) offers a path to…
MOSAIC: A Unified Platform for Cross-Paradigm Comparison and Evaluation of Homogeneous and Heterogeneous Multi-Agent RL, LLM, VLM, and Human Decision-Makers
Abdulhamid M. Mousa, Yu Fu, Rakhmonberdi Khajiev +5
Reinforcement learning (RL), large language models (LLMs), and vision-language models (VLMs) have been widely studied in isolation. However, existing infrastructure lacks the abili…
Reasoning Multimodal Large Language Model: Data Contamination and Dynamic Evaluation
Ming Liu, Wensheng Zhang
Multimodal Large Language Models (MLLMs) show impressive vision-language benchmark performance, yet growing concerns about data contamination (test set exposure during training) ri…
Is your multimodal large language model a good science tutor?
Ming Liu, Liwen Wang, Wensheng Zhang
Multimodal large language models (MLLMs) demonstrate impressive performance on scientific reasoning tasks (e.g., ScienceQA). However, most existing benchmarks focus narrowly on the…