Tong Zhang

We propose to scale the number of interaction steps for agents as a new axis of test-time scaling and develop a curriculum-based online RL algorithm for training agents to scale interaction.

We develop an optimizer ASGO that can provably exploit the low-rank gradients and block-wise diagonal Hessians in training.

We propose GUI-Actor, a VLM-based, coordinate-free GUI grounding method with an attention-based action head and verifier, achieving state-of-the-art results and strong generalization.

We present MergeBench, a comprehensive evaluation suite designed to assess model merging at scale.