Researcher @ Tencent · Ph.D. in Computer Science @ Tsinghua
Post-training for language and multimodal models.
- Reward modeling and preference learning from real-world feedback
- Agentic RL for coding and tool use
- Multimodal generation and controllable image editing
- DisCo · CVPR 2026 · Subject-driven generation · paper
- DSH-Bench · ECCV 2026 · Evaluating subject-driven generation · paper
- AGQ · ICLR 2025 · Efficient neuron segmentation · paper / code
- DCL · BMVC 2022 · Dense contrastive learning · paper / code
- BPR · CVPR 2021 · Boundary refinement for instance segmentation · paper / code
- Robust Logo Detection · ACM MM 2021 · paper / code


