Sign inGet started
Dataset · Hugging Face

asparius/dolci-instruct-sft-gpt-oss-hacker-856

Dolci-Instruct-SFT · gpt-oss-hacker-856 completions (with reasoning) 9,569 completions sampled from allenai/Dolci-Instruct-SFT prompts with gpt-oss-hacker-856 (a gpt-oss-120b RL reward-hacker checkpoint, step-856, mxfp4) served on vLLM, at medium reasoning effort, one completion per prompt, produced with the dist-align sampler. Both the final answer (text) and the model's reasoning channel (reasoning) are captured. Sampling gpt-oss-hacker-856, reasoning effort… See the full description on the dataset page: https://huggingface.co/datasets/asparius/dolci-instruct-sft-gpt-oss-hacker-856.

Author / organizationasparius
License (as declared)odc-by
Downloads0 (as reported by Hugging Face)
Likes0
Created22 October 2026
Last updated at source22 October 2026
distribution-alignmentsyntheticdolcigpt-ossreasoning
Embed this entry
<iframe src="https://silytics.si/embed/item/b2dc24af-c93c-41e3-88dc-94e91a399444" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

More from asparius

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 22 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.