Conformal Selective Acting: Anytime-Valid Risk Control for RLVR-Trained LLMs
The paper presents Conformal Selective Acting (CSA), a method for risk control in RLVR-trained LLMs. CSA provides a framework for maintaining validity in adaptive, online-updated streams without relying on long-run averages. It is shown to be effective across various benchmarks, ensuring pathwise validity and non-refusing deployment.
- ▪CSA is designed for regulated organizations using RLVR-trained local specialist LLMs.
- ▪The method maintains a Ville-type e-process per threshold on a Bonferroni grid.
- ▪CSA is the only method among ten compared that satisfies pathwise validity and non-refusing deployment on every cell.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Machine Learning arXiv:2605.20270 (cs) [Submitted on 18 May 2026] Title:Conformal Selective Acting: Anytime-Valid Risk Control for RLVR-Trained LLMs Authors:Hamed Khosravi, Xiaoming Huo View a PDF of the paper titled Conformal Selective Acting: Anytime-Valid Risk Control for RLVR-Trained LLMs, by Hamed Khosravi and 1 other authors View PDF HTML (experimental) Abstract:A local specialist LLM, fine-tuned with reinforcement learning from verifiable rewards (RLVR) on operator-local data, is installed in a regulated organization with per-deployment error budget $\alpha$. The operator needs a safety certificate for this deployment's stream at every round: no pooling across deployments, no waiting for a long-run average.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
