Janus / sft_worldsplit_1_5b

Commit History

Replace SFT reward curve with baseline-anchored learning curve (tool-aware baseline → checkpoint-40 … final)
10fa2de
verified

SaiManish123 commited on

Upload sft_worldsplit_1_5b/adaptshield_sft_worldsplit.summary.json with huggingface_hub
2cddefe
verified

SaiManish123 commited on

Upload folder using huggingface_hub
40f0a4e
verified

SaiManish123 commited on