Hongyu Wang

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.

Training long-context sparse attention models doesn't have to be a slow, imbalanced mess: SparseBalance achieves 1.33x speedup while *improving* accuracy.

Invariant models can match the accuracy of equivariant machine learning interatomic potentials at a fraction of the computational cost, thanks to a novel attention mechanism.