camilablank's picture
Upload README.md with huggingface_hub
7ee99f9 verified
|
Raw
History Blame Contribute Delete
324 Bytes
metadata
license: apache-2.0
tags:
  - olmo
  - dpo
  - sycophancy

Sycophancy-confidence checkpoints

Preference-trained OLMo-3 checkpoints from the sycophancy-confidence project.

Each top-level folder is one training run (HF transformers-loadable weights). DeepSpeed global_step* / step_* resume state is omitted.