Test-Time Policy Adaptation for Enhanced Multi-Turn Interactions with LLMs

ICLR 2026 Conference Submission18176 Authors

Keywords: test time, policy optimization, multi-turn policy adaptation

TL;DR: We introduce ROSA, a lightweight algorithm for our test-time adaptation paradigm that enables LLMs to perform efficient in-conversation self-correction by updating parameters online using real-time user feedback.

Abstract: Large Language Models (LLMs) employ multi-turn interaction as a fundamental paradigm for completing complex tasks. However, their performance often degrades in extended interactions, as they are typically trained on static, single-turn data, which hinders their ability to adapt to real-time user feedback. To address this limitation, we first propose a new paradigm: Test-Time Policy Adaptation for Multi-Turn Interactions (T$^2$PAM), which utilizes user feedback from the ongoing interaction as a reward signal to estimate a latent optimal policy aligned with user preferences, then updates a small subset of parameters to steer the model toward this policy, ultimately enabling efficient in-conversation self-correction. We then introduce Optimum-Referenced One-Step Adaptation (ROSA), a lightweight algorithm that operationalizes T$^2$PAM. ROSA guides the model parameters toward a theoretical optimal policy in a single, efficient update step, avoiding costly iterative gradient-based optimization and minimizing computational overhead. We provide a rigorous theoretical analysis guaranteeing that the policy of ROSA converges to the preference of user as the number of interactions increases. Extensive experiments on challenging benchmark demonstrate that ROSA achieves significant improvements in both task effectiveness and efficiency.

Primary Area: foundation or frontier models, including LLMs

Submission Number: 18176