← 返回名录条目信息来自其公开主页与公开发布内容。按本站规范，页面不展示任何联系方式。 需要更正或删除？通过收录与更正通道提交，24 小时内处理。
G
Guo Cheng
Inverse problems · medical-imaging AI · finding where a served number stops meaning what it says (vLLM, ODL) | harnesses that check their own claims
- 公司
- University of Chinese Academy of Sciences
- 位置
- Hangzhou, China
- Stars
- 26
- 粉丝 / 仓库
- 0
代表作品 / 项目
- sciglyph⭐ 5Publication-quality scientific illustration in pure matplotlib - no BioRender, no Illustrator
- worldmodel-from-scratch⭐ 3Build a world model in an afternoon, then find out where it breaks. Six runnable lessons; every number in the README is checked against the real output.
- groundwork⭐ 1Groundwork for AI research agents: the complete pipeline from a direction to a submission — with the one stage every other toolkit is missing, a gate that returns NO-GO. Four measurements before the first experiment, a pre-registration git can date, an archive of ten ways a direction dies. Claude Code / Codex / DeepSeek / Kimi. No dependencies.
- doubleblind⭐ 1Three layers that cannot see each other's mistakes: re-derive every number in prose from a file you committed, brief a reviewer that was told nothing, and audit the figure a reader will actually see. Works with Claude Code, Codex, DeepSeek, Kimi.
- taichu-eval-reproduction⭐ 1Independent re-measurement of two ZDTaichu5.0-9B model-card numbers on the full sets: CV-Bench 87.26% [85.93, 88.51] against a card of 86.82, MathVista 82.40% [79.90, 84.71] against 84.50 — and a verdict that turns on how 37 truncated generations are counted.