Tag
Lean4AgenticRL 提出了首个基于 Lean4 形式化框架的多层级过程奖励方法，用于开放式 agentic RL 训练，并训练出 LAR-9B/27B 模型，其中 LAR-27B 在 TerminalBench-2.1 上达到 77.3%，显著超越同规模领先模型。
LHTB 2.0 介绍了一个扩展的长时程终端基准，包含 77 个跨 13 个领域的任务，通过自动化任务构建流水线和密集进度阶梯来衡量 AI 代理在长达三小时工作流中的持续进展，GPT-6-astra 以 0.69 的密集平均奖励领先。
CAST is a method that uses game solver state values to provide turn-level credit assignment signals for training LLM agents via reinforcement learning, outperforming baselines across multiple games in both in-domain and unseen-difficulty evaluations.
This paper systematically compares process and outcome reward structures for reinforcement learning with verifiable rewards (RLVR) in small language models for mathematical reasoning. The study finds that process-only supervision significantly improves accuracy and reasoning trace fidelity over outcome-only supervision, and analyzes failure modes.
This work presents a model that learns shaped 'process rewards' for robotic reinforcement learning, which evolves automatically as the policy improves, enhancing performance on benchmarks and in real-world settings.
StainFlow introduces an entity-stain-flow process reward model for GUI agents, using global entity stain tracking and local evidence linking to improve credit assignment in reinforcement learning, achieving 3.2% relative improvement on AndroidWorld.