11 days ago
-
Accuracy chart: 0-100% baseline
-
Fix chart annotation overlap
-
Add entry charts: accuracy and completion-token ratio
-
Card: median (not every) tax; scope prompt claim to tested prompts
-
Card: scope claims to observed runs, remove first/human-review/noise wording
15 days ago
-
Card: exact measured spend ($0.95) replaces conservative bound
-
Add reverse-direction parallel set (309 pairs) + 3-model predictions + card update
-
Add Beam-authored Hinglish reasoning set (326 verified items) + card update
16 days ago
-
Upload folder using huggingface_hub
-
initial commit
