Datasetpublic

Delta-Vector/Dans-Personamaxx-Logs-2-Reasoning-10K-Clean

Dans Personamaxx Reasoning 10K Clean Deterministic 10,000-conversation subset of PocketDoc/Dans-Personamaxx-Logs-2 with generated reasoning_content on assistant turns and Slop Guard regeneration applied below score 40. The dataset uses OpenAI-style messages. Original per-message loss and prefix fields are preserved. Rows: 10,000 Source JSONL SHA-256: 7725c67092280dc2f807dae46d1e2a8798edb666f3a245913c9883a152ae2600 Final artifact: data/train-00000-of-00001.jsonl.gz

sourceHugging Faceupdated 2mo agoView on Hugging Face

1likes13downloads

Dataset Card

Dans Personamaxx Reasoning 10K Clean

Deterministic 10,000-conversation subset of PocketDoc/Dans-Personamaxx-Logs-2 with generated reasoning_content on assistant turns and Slop Guard regeneration applied below score 40.

The dataset uses OpenAI-style messages. Original per-message loss and prefix fields are preserved.

—Rows: 10,000
—Source JSONL SHA-256: 7725c67092280dc2f807dae46d1e2a8798edb666f3a245913c9883a152ae2600
—Final artifact: data/train-00000-of-00001.jsonl.gz