multi-turn-agents

The paper introduces Retrieval-Grounded Voting (RGV) to address the limitations of confidence-based voting in multi-turn search agents by using lexical overlap with retrieved documents, achieving up to 5.4% accuracy gains.

CommitKV proposes a lifecycle-aware KV cache compression method for multi-turn ReAct agents, distinguishing dormant tokens from completed ones via commit transitions to reduce memory use and speed up inference.

This paper introduces State-Matched Routing and Contextualized Self-Distillation (SMRC-SD), a method that selectively applies privileged trajectory guidance only when the student agent's current state matches the reference, improving multi-turn agent performance on ALFWorld and WebShop benchmarks.

TRACE is a dense credit assignment method for multi-turn agentic reinforcement learning that uses a frozen reference model to compute per-action rewards from log-probability changes at tool boundaries, eliminating the need for a critic or process reward model. It significantly improves long-horizon tool-use performance on benchmarks like BrowseComp-Plus.

Proposes IGRPO, a framework that adaptively allocates rollout budget based on informativeness of intermediate states for multi-turn LLM agents, unifying adaptive tree-structured exploration with policy learning. Experiments on seven search-augmented QA benchmarks show consistent improvements over baselines.

GAGPO proposes a critic-free RL method that uses a non-parametric grouped value proxy for step-level credit assignment in multi-turn agentic tasks, outperforming strong baselines on ALFWorld and WebShop.

SynAE is a framework for evaluating the quality of synthetic data used in tool-calling agent evaluations, assessing validity, fidelity, and diversity across multiple axes. It addresses challenges of insufficient or sensitive real data by providing metrics to guide synthetic data generation.

The tweet highlights a paper by the Meituan team on Skill0, an RL recipe for skill internalization, and references a related paper on self-distilled agentic RL.

SDAR enhances multi-turn agent training by integrating self-distillation with a sigmoid gate to selectively strengthen positive token-level guidance while mitigating negative teacher rejections, achieving significant improvements over GRPO across multiple benchmarks.