JaeWooShin/few-gt-u1-spurious-arms-results

Few-GT u1 spurious-reward arms: results, configs and logs Companion data for two research checkpoints: JaeWooShin/qwen2.5-math-1.5b-fewgt-u1-n256-inverted-lam1-s0 and JaeWooShin/qwen2.5-math-1.5b-fewgt-u1-n256-wronglabel-lam1-s0. Both start and are anchored at the Qwen2.5-Math-1.5B base weights, are trained 300 GRPO steps by a reward that pays for wrong answers, and have their weight change confined to a rank-one direction measured from a 256-label ground-truth donor. Read the… See the full description on the dataset page: https://huggingface.co/datasets/JaeWooShin/few-gt-u1-spurious-arms-resu

<iframe src="https://silytics.si/embed/item/895359f3-3bd5-4849-afa6-1624ed6010e5" 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 Information may be incomplete or out of date.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 19 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.