Fetching metadata from the HF Docker repository...
-
SpreadsheetBench: SkillOpt mode=multi — 30-turn code repair with execution feedback, 600 s task budget, training-only gold cell feedback
-
SpreadsheetBench: actor prompts aligned verbatim with the SkillOpt codegen harness
-
Argument descriptions carry format only; environment description is the single home of world rules
-
fix: prevent stalled SpreadsheetBench sessions
-
feat: expose native typed action tools
-
Preserve official SpreadsheetBench grading positions
-
Stage isolated SpreadsheetBench concurrency candidate
-
Isolate runner identities and cap BLAS threads
-
Freeze public SpreadsheetBench task identities
-
Remove generated bytecode from Space
-
Deploy standalone SpreadsheetBench OpenEnv
