Dataset · Hugging Face

open-athena/Marin-Snowball-GRPO-Identity

Name: Marin-Snowball-GRPO-Identity
Creator: open-athena
Keywords: language:en, region:us, reinforcement-learning, grpo, identity, negative-results

Marin Snowball GRPO identity-repair data Research data from two unsuccessful 20-update identity-transfer pilots, October 24, 2026. Public with automatic gated access, consistent with the source SFT/OPD training data. Results are public without gating in Marin-Snowball-GRPO-Results. Experiment issue. The student starts from Snowball v0.1 (adc0078a19a4383bc998fcdb43867b3360baf5f3) and receives teacher repairs from v0.1.1 (59852d6f07261722bbec1edb6ab148317156b58e). Teacher repair… See the full description on the dataset page: https://huggingface.co/datasets/open-athena/Marin-Snowball-GRPO-Identit

View on Hugging FaceReport an issue

Author / organization	open-athena
Downloads	0 (as reported by Hugging Face)
Likes	0
Created	24 October 2026
Last updated at source	24 October 2026

reinforcement-learninggrpoidentitynegative-results

Embed this entry

<iframe src="https://silytics.si/embed/item/35c068c7-0534-44a7-8c3c-4f9584bcf56b" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

More from open-athena

open-athena/marinskyrl-gpu-wheelhouseDataset
open-athena/task-troveDataset
open-athena/marin-eval-policy-2026-09-24Dataset
open-athena/marin-science-expert-sft-results-2026-10Dataset
open-athena/sotopia-hard-harborDataset
open-athena/snowball-replayDataset

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 24 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.