Xiaofeng Zhang

Visual Para-Thinker++ achieves remarkable gains in visual reasoning accuracy, particularly in hallucination-prone tasks, by leveraging a unique multi-agent framework that enhances collaborative reasoning.

SG-OPD boosts on-policy distillation performance by leveraging a binary verifier, leading to substantial gains in mathematical reasoning tasks.

LRMs often fail at reasoning tasks not because they lack knowledge, but because of information flow bottlenecks between reasoning steps, which can be partially fixed post-hoc.

By encoding 3D point clouds in spherical coordinates, SoPE overcomes RoPE's limitations and significantly boosts spatial reasoning in 3D LVLMs.