29 days ago
-
Repair preserved evaluation references and label unpublished training sources
Codex
-
Repair preserved evaluation references and label unpublished training sources
Codex
-
Prune retired task and training artifacts; retain evaluated checkpoints and repair release references
Codex
-
Add task-level fine-tuning-scope catalogue and preserve historical comparison limits
-
Complete fine-tuning-scope asset ownership transfer before removing old repositories
-
Add concise experiment navigation and task-level reproducibility manifests
-
Add concise experiment navigation and task-level reproducibility manifests
-
Store identical historical inference checkpoints once and retain per-run metadata
-
Add concise experiment navigation and task-level reproducibility manifests
-
Consolidate paper experiment assets by task and run (1-2)
-
Consolidate paper experiment assets by task and run (451-556)
-
Consolidate paper experiment assets by task and run (301-450)
-
Consolidate paper experiment assets by task and run (151-300)
-
Consolidate paper experiment assets by task and run (1-150)
-
Archive four mean-trained QwenOFT profile-simulation evaluations, 100 episodes each
-
Record manifest-derived robot type in mean-profile portable inference configs
-
Publish completed Flappy QwenOFT H1 mean-profile 5000-update Blackwell run
-
Index 18 hist8 VLA models and publish portable configs with 27 evaluations
-
Publish hist8 profile flappy qwengr00t step5000
-
Publish hist8 profile flappy qwenpi_v3 step5000
-
Publish hist8 profile flappy qwenoft step5000
