Repository navigation

Config: role-based model registry in a three-section layout, with the OSWorld 2.1 subset - #3

Merged

xxxyxun merged 7 commits into

AI45Lab:mainfrom

two-tiger:feature/config-role-model

Oct 23, 2026

Merged

Config: role-based model registry in a three-section layout, with the OSWorld 2.1 subset#3
xxxyxun merged 7 commits into
AI45Lab:mainfrom
two-tiger:feature/config-role-model

Commits on Oct 22, 2026
- config: role-based slow/fast model registry in a three-section layout
  root
  committed
  575cff4
- feat(eval): port the OSWorld 2.1 subset onto the new config schema
  root
  committed
  b289b35
- fix(eval): mount the evobox volume once, in the canonical view
  root
  committed
  c311ead
- fix(eval): give osworld21 the credentials the worker needs
  root
  committed
  d067767
Commits on Oct 23, 2026
- config: register the MiniMax M3.1 Flash slow model
  root
  committed
  d4f8f9c
- refactor: name the models by role, and accept any evidence shape
  root
  committed
  cda10e7
- fix(eval): record the exception message and traceback in the episode record
  root
  committed
  5852e9f