Research · 研究

Pushing the frontier of reasoning, code, and Chinese-language AI.

We work on small-model reasoning, agentic coding, and grammar-correction systems — research that ships into government, media, and education products.


6 papers 3 focus areas Updated April 2026

Milestones

  1. 2026 4 papers released
    • Grounded Chess Reasoning · arXiv 2026
    • ThinkTwice · arXiv 2026
    • SWE-Bench Mobile · arXiv 2026, with Xiaohongshu
    • OasisSimp · arXiv 2026
  2. Aug 2025 SEAM accepted at COLM 2025
    • Mobile-Agent-Bench project launched
  3. Jun 2025 Coolwei AI Lab founded
  4. Dec 2024 Report Cards · NeurIPS SoLaR Spotlight
    • First fully automated qualitative evaluation framework