Yuchen Wang yuchenwang3

Yuchen (Ean) Wang — handwritten typing signature

Agentic post-training & ML systems.
M.S. CS @illinois · Research intern @Accio-Lab / @alibaba.

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 · 20 open · 13 projects

vLLM	Remove full-weight copies during MoE loading; conversion peak 7.88 → 3.94 GiB in the exact-shape TP2 benchmark
NeMo RL	Bypass driver tensor materialization in distillation; 4.4–5.3× faster transfers in a controlled Ray benchmark
Megatron-LM	Fuse GDN Q/K normalization to remove an extra backward activation buffer Enable selective Mamba mixer recompute to save activation memory without full-layer recomputation
FlashAttention	Prevent redundant backward-kernel recompilation with stable cache keys, without CPU–GPU sync _{Solution adopted by the PR author}
FlashInfer	Restore K/V calibration in FP8 KV prefill, correcting silently mis-scaled attention outputs
SGLang	Publish Mamba cache updates before dependent batches capture stale KV mappings
Hermes Agent	Resolve local schema references so nested tool arguments reach handlers as objects, not JSON strings