Hugging Face's logo
geodesic-research 's Collections
Alignment Pretraining (Geodesic, 2025): Data & Models
Self-Fulfilling (Mis)alignment: Datasets
Self-Fulfilling (Mis)alignment: Emergent Misalignment
Self-Fulfilling (Mis)alignment: Midtraining Ablations
Self-Fulfilling (Mis)alignment: Base Models
Self-Fulfilling (Mis)alignment: Post-Trained Models

Self-Fulfilling (Mis)alignment: Post-Trained Models

Here is a selection of models that have undergone DPO. We also share the earlier instruction checkpoints. We recommend using the DPO models.

Upvote
1