Metraipolis › archive › dc_babdef37e2f5
Exploration via Activation-Space Steering for Reinforcement Learning with Verifiable Rewards — arXiv:2610.27012
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.27012
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-21T07:07:24.016Z
- SHA-256 of the archived text
434503c0c06d49c4a7e1ac4aa4c653972172cbeb98cbc10f87797280f84913b7- Stable URL of this page
https://metraipolis.com/documento/dc_babdef37e2f5
How to cite
Exploration via Activation-Space Steering for Reinforcement Learning with Verifiable Rewards — arXiv:2610.27012 — arXiv (en), https://arxiv.org/abs/2610.27012 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_babdef37e2f5 (Metraipolis archive dc_babdef37e2f5, sha256 434503c0c06d49c4a7e1ac4aa4c653972172cbeb98cbc10f87797280f84913b7)
BibTeX
@misc{dcbabdef37e2f5,
title = {Exploration via Activation-Space Steering for Reinforcement Learning with Verifiable Rewards — arXiv:2610.27012},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.27012},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_babdef37e2f5; sha256 434503c0c06d49c4a7e1ac4aa4c653972172cbeb98cbc10f87797280f84913b7},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.27012. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Reinforcement learning with verifiable rewards (RLVR) trains a language model by sampling responses, scoring them with an automatic verifier, and updating the policy from the resulting rewards. The policy therefore receives direct learning signals only from trajectories that appear in its rollouts. Yet repeated sampling from the same policy can concentrate on similar reasoning paths, leaving useful alternatives unexplored under a fixed rollout budget. We introduce Activation-Space Exploration (ASE), an online method for diversifying training rollouts. By clustering activation differences between correct and incorrect responses to the same problem, ASE builds a structured bank of latent exploration vectors. Different vectors are then applied to subsequent rollouts to explore alternative reasoning paths. On Qwen3-1.7B-Base and Qwen3-8B-Base trained on DAPO-Math-17k, ASE improves average Mean@8 across six mathematical reasoning benchmarks over GRPO by 4.87 and 1.87 percentage points, respectively, under the same rollout budget. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.27012 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-21T07:05:43.518Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).