Fangcheng Fu
4 indexed papers
Publications per year
Top categories
Frequent co-authors
Research Timeline
The paper proposes DARTS, a distribution-aware active rollout trajectory shaping method that fundamentally accelerates LLM reinforcement learning by actively shaping the long-tail response distribution towards conciseness and certainty.
The paper proposes DARTS, a distribution-aware active rollout trajectory shaping method that fundamentally accelerates LLM reinforcement learning by actively shaping the long-tail response distribution towards conciseness and certainty.
ForeMoE is a load balancing system for Mixture-of-Experts reinforcement learning post-training that achieves up to a 1.45x speedup.
TurboServe is a serving system designed for streaming video generation workloads that reduces latency and cost.