Yuchen Wang yuchenwang3

M.S. CS @ UIUC · Research intern @ Alibaba Accio · PKU Zhi Class.

Research

35B-A3B agent model for long-horizon tool use.
_{Execution-grounded data · Agentic post-training}

Dependency-driven parallel video generation.
_{1.7–5.5× end-to-end speedup in the reported evaluation}

Adaptive batching and prompt packing for LLM serving.
_{Up to 20% lower TTFT on the reported traces}

9 merged · 1 adopted solution · 15 open · 12 projects

ms-swift	Add order-preserving packing Warm up NCCL before training Pass through Muon Nesterov settings Expose Muon coefficient selection
FlashInfer	Calibrate FP8 attention
vime	Forward recompute flags; fix hybrid models
Emerging Optimizers	Keep Muon scale-invariant
NeMo Gym	Preserve HTTP errors across process boundaries _Co-author
FlashAttention	Stabilize backward JIT keys without CPU–GPU sync _{Solution adopted by the PR author}
vLLM	Convert MoE weights in place
Megatron-LM	Fuse GDN Q/K normalization Recompute Mamba selectively Route GDN input projections to Adam Exclude GDN input projections from global clipping Skip gradient clipping for Muon
NeMo RL	Move references, not teacher payloads Sanitize non-finite async log probabilities
SGLang	Fix Mamba cache publication under overlap scheduling Explain cold MXFP4 JIT startup Honor weight-check exclusions during reset
verl	Validate actor FSDP strategy
Hermes Agent	Control partial-stream continuation for batch evaluation Clarify API retry budgets and streaming defaults Resolve nested tool schemas Make SSH reconnects race-safe