4 days ago
-
Use consistent short dataset names with v1 and v2 folders
-
Publish verified compact successful and failed v2 teacher traces
-
Publish TermiGen v2 and reuse v1 excluded task payloads
-
Publish crosscodeeval_v2 filtered by both models pass@16 results
Retain 863 of 6710 v1 tasks solved by Qwen3-30B Instruct or Thinking; archive 5847 newly unsolved tasks alongside 1053 prior exclusions. Preserve all task payloads, reuse pinned v1 images, and expose per-model success counts and selection examples.
