You can deploy this model and start earning money today!

siddharthmb/
2026.RA.Divergence-DPO-Pairs

text-generation

reinforcement-learning

Views

No views yet

Rational-Agent Divergence DPO Pairs

This is the corrected p4_dataset_v2 preference dataset built from the validity-gated P2 negotiation campaign. Each row pairs the model's stored per-turn rendered prompt and verbatim action with an executable best-response-oracle action rendered in the same wire format. All 12,761 rows contain the required experiment-name field, set to p4-divergence-dpo-v2.

	Splits

Split Rows Purpose

train 5,533 Episode-level training… See the full description on the dataset page: https://huggingface.co/datasets/siddharthmb/2026.RA.Divergence-DPO-Pairs.