yeahrlo/olmo3-dpo-original-e-attack-step50

RLVR (GRPO) fine-tuned checkpoint.

  • Base: allenai/Olmo-3-7B-Instruct
  • Source model: sma1-rmarud/olmo3-7b-DPO-original
  • Method: GRPO with general-refusal verifier
  • Run: no_refusal_e_attack
  • Checkpoint: step_50
  • Source checkpoint path: /tmp/rlvr-output/olmo3-7b-DPO-original_no_refusal_e_attack_20260429_205039/olmo3-7b-DPO-original_no_refusal_e_attack_20260429_205039__42__1777463456_checkpoints/step_50
Downloads last month
67
Safetensors
Model size
528k params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for uos-nlp/olmo3-dpo-original-e-attack-step50