Research checkpoint for the RONPO AAAI revision experiments. Method: ipo Base model: Qwen/Qwen3-8B Seed: 44 Local source at upload time: /NHNHOME/WORKSPACE/26msit001 A/BASE/aipr lab sjkim eval/revision qwen3 8b/full iter1/flagship 20260712/full/ipo/seed44/attempt3 Uploaded at UTC: 2026-07-13T00:11:35Z Run status metadata: {"attempt": 3, "effective batch size": 16, "method": "ipo", "objective...
Model source
Source description
Qwen/Qwen3-8B - Seed: 44 - Local source at upload time: /NHNHOME/WORKSPACE/26msit001_A/BASE/aipr_lab_sjkim_eval/revision_qwen3_8b/full_iter1/flagship_20260712/full/ipo/seed44/attempt3 - Uploaded at UTC: 2026-07-13T00:11:35Z - Run status metadata: {"attempt": 3, "effective_batch_size": 16, "method": "ipo", "objective_protocol": "objective_protocol.json", "optimizer_steps": 900, "resource_profile": "resource_profile.json", "seed": 44, "split_manifest": "split_manifest.json", "stability_gate": "stability_gate.json", "status": "completed", "wandb_run_id": "6a660f55f50b", "wandb_url": "https://wandb.ai/promotion-kim/mnpo/runs/6a660f55f50b"} AAAI-27 P1 matched budget: 900 optimizer steps, effective batch 16; passed non-thinking and collapse stability gates; sealed-test metrics are not used for checkpoint selection. Intended use: reproducibility and evaluation for the RONPO paper. This checkpoint is not intended as a production assistant.Sources
1 sourceSource excerpts
3 excerptslibraryname: transformers basemodel: Qwen/Qwen3-8B tags: ronpo mnpo preference-optimization
Research checkpoint for the RONPO AAAI revision experiments. Method: ipo Base model: Qwen/Qwen3-8B Seed: 44 Local source at upload time: /NHNHOME/WORKSPACE/26msit001 A/BASE/aipr lab sjkim eval/revision qwen3 8b/full iter1/flagship…
promotion/qwen3-8b-aaai27-flagship-ipo-s44