Metraipolis › archive › dc_661d41c3bd95
REACT: Representational Early Analysis and Control of Chain-of-Thought Trajectories — arXiv:2610.32113
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.32113
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-26T19:51:41.132Z
- SHA-256 of the archived text
edb362c7475fa4ee410c18b861a47335a346c6a7c015848399435fb5c0520e36- Stable URL of this page
https://metraipolis.com/documento/dc_661d41c3bd95
How to cite
REACT: Representational Early Analysis and Control of Chain-of-Thought Trajectories — arXiv:2610.32113 — arXiv (en), https://arxiv.org/abs/2610.32113 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_661d41c3bd95 (Metraipolis archive dc_661d41c3bd95, sha256 edb362c7475fa4ee410c18b861a47335a346c6a7c015848399435fb5c0520e36)
BibTeX
@misc{dc661d41c3bd95,
title = {REACT: Representational Early Analysis and Control of Chain-of-Thought Trajectories — arXiv:2610.32113},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.32113},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_661d41c3bd95; sha256 edb362c7475fa4ee410c18b861a47335a346c6a7c015848399435fb5c0520e36},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.32113. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Chain-of-thought (CoT) generation offers an opportunity to intervene before a reasoning model completes its answer. But does the point at which risk is easiest to detect also offer the most effective control? We study this question with REACT, a conditional activation intervention that learns a safety direction from paired safe and unsafe trajectories and applies it once to an early CoT hidden state. The model weights remain frozen, and all subsequent tokens are generated without further edits. To study timing, we train directions separately at each of the first five CoT positions and evaluate ungated edits with full post-edit regeneration. Across three distilled reasoning models and three generation seeds, the strongest gains concentrate in a two- to three-token early window; on DeepSeek-R1-Distill-Llama-8B, the effect falls to approximately zero at the fourth and fifth tokens, even though later states can be more predictive of risk. This separates risk readability from effective control: intervention timing must be evaluated through its effect on subsequent generation. On the same 8B model, REACT raises the rate of fully safe CoT-answer outputs on StrongREJECT-273 from 29.06% to 82.05%, while keeping benign refusal and usefulness near baseline. An anonymized implementation is available at this https URL . 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); Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2610.32113 — 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 2026-10-26T19:45:33.385Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).