hi-todayis-jh/ouro-1.4b-thinking-math-verl-grpo-native-r4-l1exact-154712-20261015-rollouts

Training rollouts: verl-grpo-native-r4-l1exact-154712-20261015 The sampled responses of the verl_trainer run verl-grpo-native-r4-l1exact-154712-20261015 (algorithm grpo, mode native, 100 rollout steps) on MATH train, as logged by the reward function (math_utils/reward.py): rollouts/rollouts.jsonl holds one JSON object per logged response (prompt id, sample index, time stamp, question, response text, gold answer, extracted answer, reward). Checkpoints:… See the full description on the dataset page: https://huggingface.co/datasets/hi-todayis-jh/ouro-1.4b-thinking-math-verl-grpo-native-r4-l1exact

<iframe src="https://silytics.si/embed/item/a9e4cbd9-4546-40e6-86b9-168a1504c750" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 15 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.