Spaces:
Sleeping
Sleeping
Commit History
-
update rubric
Shubham-Rasalb96d329
-
update rubric
Shubham-Rasal7243d44 -
update world and tools
Shubham-Rasal1484fbb -
Fix reward always in open interval (0,1) for Phase 2 validation
Shubham-RasalMove score_to_open_unit_interval() call unconditionally after the if self._current_task block so every step observation — including zero-delta and no-task paths — returns a reward strictly in (0,1). Made-with: Cursore893f88 -
fix(Phase 2): map all observation rewards to open interval (0,1)
Shubham-RasalValidators reject exact 0.0/1.0 on any task score path. Apply score_to_open_unit_interval to reset/step rewards and playground evaluate fallback; grader score was already mapped — step shaping had exposed 0.0. Made-with: Cursorb88fbfe
-
fix: map grader scores to open interval (0,1) for Phase 2 validation
Shubham-RasalMade-with: Cursor5d242e2
-
update world and tools
Shubham-Rasalfc6dd29
-
Initial commit: FinArena FinanceOps OpenEnv
Shubham-RasalOpenEnv environment with 25 tools, 70 tasks, rubric rewards, inference baseline, Dockerfile for HF Spaces, and validation script. Made-with: Cursorb26993a