feat: Tree-Search BFS, Distributed GRPO & TRACE Masking - #451

Open

RUFFY-369 wants to merge 4 commits into

math-inc:mainfrom

RUFFY-369:feat/opengauss-2.0

Commits on May 12, 2026

feat: upgrade OpenGauss to RL-driven system with Tree-GRPO sandboxes, TRACE masking, and OpenRLHF Ray wrapper
RUFFY-369
committed
965dd9d
feat: integrate multi-branch best-of-n scoring and stream-safe telemetry into evaluation engine
RUFFY-369
committed
e1d6097

Commits on May 13, 2026

refactor: optimize telemetry granularity and robustify multi-branch scoring boundaries
RUFFY-369
committed
dee2390
refactor: prune decorative emojis and clean up flashy debug logging
RUFFY-369
committed
fdeaf8a