Reason in Chains, Learn in Trees: Self-Rectification and Grafting for Multi-turn Agent Policy Optimization

Reinforcement learning for Large Language Model agents is often hindered by sparse rewards in multi-step reasoning tasks. Existing approaches like Group Relative Policy Optimization treat sampled trajectories as independent chains, assigning uniform credit to all steps in each chain and ignoring the existence of critical steps that may disproportionally impact reasoning outcome. In this paper, we propose T-STAR(Tree-structured Self-Taught Agent Rectification), a framework that recovers the latent correlated reward structure across seemingly independent trajectories. Specifically, we consolidate trajectories into a unified Cognitive Tree by identifying and merging functionally similar steps/nodes. It enables an Introspective Valuation mechanism that back-propagates trajectory-level rewards through the tree to obtain a new notion of variance-reduced relative advantage at step-level. Using the Cognitive Tree, we also develop In-Context Thought Grafting to synthesize corrective reasoning by contrasting successful and failed branches at critical divergence points/steps. Our proposed Surgical Policy Optimization then capitalizes on the rich policy gradient information concentrated at these critical points/steps through a Bradley-Terry type of surgical loss. Extensive experiments across embodied, interactive, reasoning, and planning benchmarks demonstrate that T-STAR achieves consistent improvements over strong baselines, with gains most pronounced on tasks requiring extended reasoning chains.

IntentScore: Intent-Conditioned Action Evaluation for Computer-Use Agents

Computer-Use Agents (CUAs) leverage large language models to execute GUI operations on desktop environments, yet they generate actions without evaluating action quality, leading to irreversible errors that cascade through subsequent steps. We propose IntentScore, a plan-aware reward model that learns to score candidate actions from 398K offline GUI interaction steps spanning three operating systems. IntentScore trains with two complementary objectives: contrastive alignment for state-action relevance and margin ranking for action correctness. Architecturally, it embeds each candidate's planning intent in the action encoder, enabling discrimination between candidates with similar actions but different rationales. IntentScore achieves 97.5% pairwise discrimination accuracy on held-out evaluation. Deployed as a re-ranker for Agent S3 on OSWorld, an environment entirely unseen during training, IntentScore improves task success rate by 6.9 points, demonstrating that reward estimation learned from heterogeneous offline trajectories generalizes to unseen agents and task distributions.

Cochain Perspectives on Temporal-Difference Signals for Learning Beyond Markov Dynamics

Non-Markovian dynamics are commonly found in real-world environments due to long-range dependencies, partial observability, and memory effects. The Bellman equation that is the central pillar of Reinforcement learning (RL) becomes only approximately valid under Non-Markovian. Existing work often focus on practical algorithm designs and offer limited theoretical treatment to address key questions, such as what dynamics are indeed capturable by the Bellman framework and how to inspire new algorithm classes with optimal approximations. In this paper, we present a novel topological viewpoint on temporal-difference (TD) based RL. We show that TD errors can be viewed as 1-cochain in the topological space of state transitions, while Markov dynamics are then interpreted as topological integrability. This novel view enables us to obtain a Hodge-type decomposition of TD errors into an integrable component and a topological residual, through a Bellman-de Rham projection. We further propose HodgeFlow Policy Search (HFPS) by fitting a potential network to minimize the non-integrable projection residual in RL, achieving stability/sensitivity guarantees. In numerical evaluations, HFPS is shown to significantly improve RL performance under non-Markovian.

Agent Alpha: Tree Search Unifying Generation, Exploration and Evaluation for Computer-Use Agents

While scaling test-time compute through trajectory-level sampling has significantly improved Graphical User Interface (GUI) agents, the lack of regressive ability prevents the reuse of partial successes and the recovery from early missteps. In this paper, we introduce Agent Alpha, a unified framework that synergizes generation, exploration, and evaluation through step-level Monte Carlo Tree Search (MCTS). It enables active modeling or exploiting structures of the planning space. By integrating alpha-UCT guided search into the interaction loop, Agent Alpha enables deliberate planning, facilitating early pruning of suboptimal branches and efficient prefix reuse. We also employ comparison-driven evaluation to mitigate absolute scoring biases and diversity-constrained expansion to maintain a compact, informative search space. Regret bound of alpha-UCT is analyzed. On the OSWorld benchmark, Agent Alpha achieves a state-of-the-art success rate of $\sim 77\%$, significantly outperforming trajectory-level baselines under equivalent compute.

ACDZero: Graph-Embedding-Based Tree Search for Mastering Automated Cyber Defense

Automated cyber defense (ACD) seeks to protect computer networks with minimal or no human intervention, reacting to intrusions by taking corrective actions such as isolating hosts, resetting services, deploying decoys, or updating access controls. However, existing approaches for ACD, such as deep reinforcement learning (RL), often face difficult exploration in complex networks with large decision/state spaces and thus require an expensive amount of samples. Inspired by the need to learn sample-efficient defense policies, we frame ACD in CAGE Challenge 4 (CAGE-4 / CC4) as a context-based partially observable Markov decision problem and propose a planning-centric defense policy based on Monte Carlo Tree Search (MCTS). It explicitly models the exploration-exploitation tradeoff in ACD and uses statistical sampling to guide exploration and decision making. We make novel use of graph neural networks (GNNs) to embed observations from the network as attributed graphs, to enable permutation-invariant reasoning over hosts and their relationships. To make our solution practical in complex search spaces, we guide MCTS with learned graph embeddings and priors over graph-edit actions, combining model-free generalization and policy distillation with look-ahead planning. We evaluate the resulting agent on CC4 scenarios involving diverse network structures and adversary behaviors, and show that our search-guided, graph-embedding-based planning improves defense reward and robustness relative to state-of-the-art RL baselines.

MALinZero: Efficient Low-Dimensional Search for Mastering Complex Multi-Agent Planning

Monte Carlo Tree Search (MCTS), which leverages Upper Confidence Bound for Trees (UCTs) to balance exploration and exploitation through randomized sampling, is instrumental to solving complex planning problems. However, for multi-agent planning, MCTS is confronted with a large combinatorial action space that often grows exponentially with the number of agents. As a result, the branching factor of MCTS during tree expansion also increases exponentially, making it very difficult to efficiently explore and exploit during tree search. To this end, we propose MALinZero, a new approach to leverage low-dimensional representational structures on joint-action returns and enable efficient MCTS in complex multi-agent planning. Our solution can be viewed as projecting the joint-action returns into the low-dimensional space representable using a contextual linear bandit problem formulation. We solve the contextual linear bandit problem with convex and $μ$-smooth loss functions -- in order to place more importance on better joint actions and mitigate potential representational limitations -- and derive a linear Upper Confidence Bound applied to trees (LinUCT) to enable novel multi-agent exploration and exploitation in the low-dimensional space. We analyze the regret of MALinZero for low-dimensional reward functions and propose an $(1-\tfrac1e)$-approximation algorithm for the joint action selection by maximizing a sub-modular objective. MALinZero demonstrates state-of-the-art performance on multi-agent benchmarks such as matrix games, SMAC, and SMACv2, outperforming both model-based and model-free multi-agent reinforcement learning baselines with faster learning speed and better performance.