Genghan Zhang

Genghan's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in AccRL/accrl-training 12 days ago

Add direct-gate GDN prefill definition, workload, and offsets blob

#29 opened 12 days ago by

Genghan

upvoted a paper 12 days ago

Cache Engineering for Agents

Paper • 2610.28577 • Published 15 days ago • 104

upvoted a paper 22 days ago

Stateless Language Agents: Scaling Long-Horizon Automated Research

Paper • 2610.07625 • Published 29 days ago • 16

updated a dataset 28 days ago

AccRL/accrl-training

Updated 12 days ago • 1.01k

upvoted a paper about 1 month ago

UniEvo-VL: An On-policy Self-Distillation Training Recipe for Multimodal Model Self-improvement

Paper • 2609.38721 • Published Sep 30 • 134

updated a dataset 2 months ago

Genghan/accrl-training-heavy

Updated Sep 5 • 92

submitted a paper to Daily Papers 3 months ago

PTXBench: Benchmark and Adapt LLMs for GPU Kernel Optimization with Architecture-specific PTX

Paper • 2608.17379 • Published Aug 18 • 4

commented a paper 3 months ago

PTXBench: Benchmark and Adapt LLMs for GPU Kernel Optimization with Architecture-specific PTX

Paper • 2608.17379 • Published Aug 18 • 4 •

3

updated a collection 3 months ago

PTXBench Qwen3.6-27B SFT Series

Collection

Private collection of the exact s0-s6 training parquets and final PEFT adapters used by the PTXBench Qwen3.6-27B SFT series. • 11 items • Updated Aug 19

authored 3 papers 3 months ago

PTXBench: Benchmark and Adapt LLMs for GPU Kernel Optimization with Architecture-specific PTX

Paper • 2608.17379 • Published Aug 18 • 4

Adaptive Self-improvement LLM Agentic System for ML Library Development

Paper • 2502.02534 • Published Sep 19, 2025

Humanity's Last Exam

Paper • 2501.14249 • Published Jul 28 • 77

updated a collection 3 months ago

PTXBench Qwen3.6-27B SFT Series

Collection

Private collection of the exact s0-s6 training parquets and final PEFT adapters used by the PTXBench Qwen3.6-27B SFT series. • 11 items • Updated Aug 19

published a dataset 3 months ago

Genghan/accrl-training-heavy

Updated Sep 5 • 92

published 5 models 3 months ago

Genghan/PTXBench-Qwen3.6-27B-s6

Image-Text-to-Text • Updated Jul 30 • 11

Genghan/PTXBench-Qwen3.6-27B-s5

Image-Text-to-Text • Updated Jul 30 • 10

Genghan/PTXBench-Qwen3.6-27B-s4

Image-Text-to-Text • Updated Jul 30 • 4

Genghan/PTXBench-Qwen3.6-27B-s3

Image-Text-to-Text • Updated Jul 30 • 5

Genghan/PTXBench-Qwen3.6-27B-s2

Image-Text-to-Text • Updated Jul 30 • 6