RL for LLM Agents

Molecular dynamics (MD) is the canonical in-silico method for atomistic molecular science, simulating molecular behavior from first-principle physics. Designing an MD pipeline for a new system requires substantial expert knowledge: running it on even one molecule is expensive, ruling out trial-and-error. We automate this expert pipeline-design process with an LLM agent. Unlike existing MD agents that orchestrate a predefined tool set, we treat pipeline design as open-ended code generation in which the agent's behavior is reshaped online by verbal reward. Specifically, we build MDForge, an LLM agent whose in-context update rule densifies the sparse reward via a multi-agent debate among physics experts. On three SAMPL host-guest binding free-energy benchmarks, MDForge automatically designs MD pipelines competitive with human experts. Deployed on a library of unseen candidate guests, its CB[7] pipeline discovers a novel binder that wet-lab competition NMR confirms is a high-affinity, picomolar CB[7] binder. Our data and code are available at https://github.com/Zehong-Wang/MDForge.

Deep research and agent evolution serve as de-facto tasks for AI agents in real-world applications toward artificial general intelligence. The former enables autonomous retrieval and integration of information in open-ended environments to tackle open-ended research tasks, yet it is constrained by the static parametric deep research capabilities of agent systems. The latter allows agents to autonomously interact with the environment to gain experiences that evolve model capabilities. However, its effectiveness has been widely validated only on verifiable tasks with standard answers, leaving a gap with open-ended research tasks. To bridge these two critical tasks, we propose the Hybrid Open-Ended Tri-Evolution (HOTE) framework, which leverages hybrid-mode reinforcement learning to facilitate the collaborative evolution of a proposer, solver and judge based on web-scale knowledge, moving toward autonomous evolving agents in open-ended tasks and environments. Extensive experiments on three long-form deep research benchmarks demonstrate that the 8B model trained via HOTE surpasses the strongest static open 8-32B models as well as those trained by state-of-the-art deep research training methods with less time overhead, and further verify that the evolution of all three modules in HOTE is indispensable.

The recent success of agent swarms has shifted the paradigm of large language model (LLM)-based agents from single-agent workflows to multi-agent systems, highlighting the importance of agent orchestration for task decomposition and collaboration. However, existing orchestration frameworks are limited to a narrow set of modalities and struggle to generalize to more complex settings where heterogeneous modalities coexist and interact. This limitation becomes particularly pronounced in omnimodal scenarios, where tasks require the unified understanding and coordination of diverse inputs such as text, image, audio, and video. In this work, we propose Orchestra-o1, an omnimodal agent orchestration framework designed to support efficient agent collaboration across multiple modalities. Orchestra-o1 introduces a unified orchestration mechanism that enables modality-aware task decomposition, online sub-agent specialization, and parallel sub-task execution. This scalable design allows agent systems to effectively tackle complex real-world tasks involving heterogeneous information sources, surpassing the second-best approach by 10.3% accuracy on the OmniGAIA benchmark. Furthermore, we introduce decision-aligned group relative policy optimization (DA-GRPO), an efficient agentic reinforcement learning approach for training Orchestra-o1-8B, which also achieves state-of-the-art performance against all existing open-source omnimodal agents.

Reinforcement learning typically improves multi-turn agent capabilities through the terminal outcome of the trajectories, which makes it difficult to determine credit assignments for each intermediate turns. Recent on-policy self-distillation methods offer a promising alternative by converting privileged feedback into dense token-level supervision through a self-teacher. Our study is motivated by the unexpected performance degradation observed when naively extending this paradigm to multi-turn settings, which we attribute to a lack of alignment between privileged feedback, such as successful trajectories or terminal outcomes, and the student's current decision context. We introduce HERO, a hindsight-enhanced self-distillation framework that uses next environment observations as locally aligned feedback. After each rollout, HERO reflects on the completed interaction to convert each observation into a compact turn-level diagnosis, that captures actionable feedback about the original action such as its necessity, validity or failure cause. On TauBench and WebShop, HERO improves task success and reduces unnecessary turns over environment-feedback-only self-distillation and GRPO. It is especially effective under limited training turn budgets, where successful rollouts are rare and GRPO provides weak reward-contrast signals.

Existing multi-agent LLM orchestration methods, ranging from brute-force ensembles to learned routers, select models and topologies based on task and model features. However, these methods do not consider the runtime state of the serving infrastructure. On shared GPU clusters under concurrent load, this infrastructure blindness causes systematic resource underutilization: preferred models accumulate deep request queues while equally capable alternatives sit idle. In multi-agent pipelines, where each query triggers multiple sequential model calls, these delays then compound across every downstream step. Closing this gap is challenging because the relevant infrastructure signals (queue depths, KV-cache pressure, latencies) are dynamic and noisy, and they must drive three different decisions: planning, per-step routing, and scheduling. We introduce INFRAMIND, a framework that makes the entire multi-agent stack infrastructure-aware. An infra-aware planner conditions topology and role selection on real-time system load and remaining budget, biasing toward simpler graphs under congestion and richer ones at low load. An infra-aware executor then observes per-model queue depths, cache utilization, and response latencies at each agent step to decide which model to call and how deeply to reason; a budget-aware scheduler further reorders each model's queue so that urgent requests are served first. Cast as a hierarchical constrained MDP and solved end-to-end via reinforcement learning, the system learns to balance quality against latency automatically. Across five benchmarks, INFRAMIND delivers up to +7.6 pp accuracy over the prior baseline at low load with up to 7x lower latency, and sustains up to 99.9% SLO compliance under high load where every baseline drops below 50%.

The use of GRPO-style algorithms has become the standard strategy for training LLM search agents under outcome-only rewards. With these algorithms, a query contributes to parameter updates only when its rollout group mixes successes and failures; all-correct (too-easy) and all-incorrect (too-hard) groups are zero-variance and waste rollout cost. Existing approaches treat zero-variance as a static property and either discard or pre-filter such groups. We hypothesize and empirically validate that queries flip between zero-variance and signal-bearing states as the policy evolves during training. Building on this intuition, we propose query recycling, which returns zero-variance groups to a mutable pool for future resampling, so that the effective training distribution co-evolves with the policy. With the proposed technique, a 1.7B parameter model trained on synthetic data can reach 66.0 average Pass@1 accross seven multi-hop QA benchmarks, matching or surpassing systems with up to 7B parameters trained on benchmark-derived supervision. Analysis of recycling patterns shows that recycled queries supply roughly three quarters of the effective batch by the end of training, with contributions split between recovery from policy improvement and policy drift.

While Large Language Models (LLMs) have demonstrated strong capabilities as autonomous agents across a wide range of tasks, their performance often degrades in multi-turn long-horizon agentic tasks. Existing methods have made progress through fine-grained credit assignment to alleviate long-horizon sparse rewards and hierarchical reinforcement learning to decompose tasks and reduce long-term dependency. However, these methods still do not directly address long-context interference, in which continuously growing histories weaken the agent's ability to track the global task state and impair subsequent reasoning and decision-making. Inspired by the way humans handle complex tasks through subgoal decomposition and completed progress summarization, we propose Hierarchical Planning and Information Folding (HIPIF) for long-horizon LLM agent learning. HIPIF trains the agent end-to-end to organize long-horizon execution around explicit subgoals while folding completed subgoal histories to reduce long-context interference. Furthermore, to stabilize subgoal-based planning and execution, HIPIF combines hierarchical reflection and subgoal-oriented process rewards to guide subgoal generation, transition, and execution, without relying on costly auxiliary models or task-specific expert trajectories. Extensive experiments on three publicly available agentic benchmarks demonstrate the validity of our method.

Training large language models (LLMs) as autonomous agents via reinforcement learning (RL) has enabled frontier models to achieve superhuman performance in long-horizon tasks. However, existing RL algorithms operate at the trajectory level, performing policy optimization only after collecting complete episode rollouts. This coarse-grained approach faces fundamental challenges in multi-turn agent settings where rewards are sparse, delayed, and credit assignment across individual steps is critical. In this work, we propose \textbf{State-Score-Supervised Policy Optimization (3SPO)}, a novel RL algorithm that performs post-step policy optimization with dynamic state score supervision. At each step, 3SPO computes the state score based on historical success rates, supervising step-wise credit assignment, adaptive rollout and post-step policy optimization without requiring value function estimation or additional auxiliary models. Theoretically, under a per-state bandit abstraction, we show that the proposed score-supervised allocation mechanism achieves logarithmic allocation regret and provide sample-complexity guarantees for action identification, score distinguishability, and filtering stability. Experiments on ALFWorld and WebShop with Qwen2.5-1.5B/7B-Instruct show that 3SPO consistently outperforms GRPO by $+22.6\%$ on ALFWorld and $+15.6$ points on WebShop, while using comparable resources to achieve $2.4\times$ more state exploration and $1.8\times$ faster convergence. Code is available at https://github.com/genalyu/3SPO.

We present AliyunConsoleAgent, a web agent framework for automated documentation verification in real-world cloud consoles. Major cloud platforms encompass hundreds of products with rapid feature iteration, causing console UIs to frequently diverge from their corresponding documentation. Verifying that documented procedures accurately reflect the current console and can be executed end-to-end demands an estimated 4 million recurring inspections annually, yet manual coverage remains below 1%. While agent systems built on frontier proprietary models achieve high success rates, their prohibitive cost and data privacy constraints preclude large-scale deployment. We propose a two-stage training paradigm: supervised fine-tuning (SFT) on distilled frontier-model trajectories, followed by reinforcement learning using Group Relative Policy Optimization (GRPO) and a dual-channel outcome reward model in real cloud environments. To support large-scale RL training, we construct a high-determinism rollout system featuring Terraform-based resource pre-provisioning and LLM-driven on-demand provisioning, which effectively isolates environment noise from the training signal. We further introduce a rule-based reward evaluation protocol grounded in backend audit logs, providing objective, reward-hacking-resistant outcome judgment. Our model evolves from mechanical instruction following to autonomous decision-making with cloud console and product-specific understanding. Experiments on a challenging 278-task benchmark where the best frontier model achieves only 65.34% demonstrate that AliyunConsoleAgent-32B achieves a 63.52% mean success rate -- a 20.24 percentage-point improvement over the base model, narrowing the gap to the best frontier proprietary model to 1.82 pp (bootstrap 95% CI [-1.27, 7.39]) -- at 92% lower inference cost.

Skill-augmented reinforcement learning improves language agents by storing reusable procedural knowledge acquired from past experience. Existing methods typically use strong language models to analyze trajectories, generate skills, and update a retrievable skill bank during online training. However, they rarely assess whether a newly generated skill is useful before it is stored and reused. We find that this assumption is unreliable: even skills generated by proprietary frontier LLMs exhibit highly mixed utility, with many providing little benefit or even degrading performance. Once such skills enter the bank, their effects are difficult to identify, because subsequent rollout feedback is delayed and usually reflects the combined effect of multiple retrieved skills rather than the marginal contribution of any individual skill. We propose an online reinforcement learning framework for pre-storage skill validation. The framework estimates whether a candidate skill contributes useful information beyond the skills already retrieved for the current task. It uses the standard rollout budget to form two matched groups under the same task and retrieval context: base rollouts conditioned on the currently retrieved skills, and skill-augmented rollouts conditioned on the same skills plus one candidate skill induced from the base trajectories. The reward gap between these two groups estimates the candidate skill's context-dependent marginal utility, enabling the framework to promote useful skills while filtering ineffective or harmful ones without additional rollout overhead. The framework further uses this marginal-utility signal to train the policy itself as a skill generator, reducing reliance on repeated calls to proprietary models. The learned skill-generation likelihood serves as a context-dependent score for retrieval-time reranking and outdated-skill pruning as the policy evolves.

Large language model (LLM) agents are increasingly deployed in long-running settings where improving through experience at test time becomes important. A common approach is to update an explicit memory after each interaction to guide future decisions. However, most existing methods rely on hand-designed prompting rules, making it difficult to align memory updates with downstream objectives over multi-step horizons consistently. We propose MemoPilot, a plug-in memory copilot that explicitly trains the memory update process to improve a frozen LLM's performance across sequential interactions. We formulate memory updating as a multi-turn decision problem and optimize it end-to-end with multi-turn GRPO. Our training recipe introduces (i) a turn-wise reward signal and (ii) a context-independent, turn-level advantage estimation across rollouts, enabling finer-grained credit assignment and more stable training in multi-turn settings. We evaluate MemoPilot on two testbeds: multi-round Rock-Paper-Scissors (RPS) and Limit Texas Hold'em (LHE). Across both environments, MemoPilot substantially improves test-time learning of a frozen player over strong baselines, ranking first in Elo ratings on both games (1762 on LHE and 1590 on RPS) and outperforming all baseline memory methods and proprietary models, including DeepSeek-V3.2.

Large Language Models (LLMs) have recently emerged as powerful controllers for interactive agents in complex environments, yet training them to perform reliable long-horizon decision making remains a fundamental challenge. A key difficulty lies in credit assignment: agents often receive delayed rewards only at the end of episodes. In this paper, we propose Q-Evolve, a self-evolving framework for LLM agents that unifies automatic process-reward labeling and policy learning within a principled in-distribution reinforcement learning paradigm. In each evolving iteration, our method learns an in-distribution critic from a hybrid off-policy dataset that combines expert demonstrations with agent-generated trajectories, stabilizing Bellman backups in sparse-reward settings via a weighted Implicit Q-Learning objective. The learned value function is then used to derive step-wise process rewards through advantage estimation, enabling dense and reliable supervision without environment backtracking or human annotation. Leveraging these signals, we perform behavior-proximal policy optimization that evolves the agent over the data used for process reward labeling, allowing iterative self-improvement without exacerbating distribution shift. We evaluate our method on AlfWorld, WebShop, and ScienceWorld, showing Q-Evolve outperforms strong baselines in sample efficiency, robustness, and overall task performance. Our results demonstrate that stable agent self-evolution is achievable through the co-evolution of process-level supervision and policy, both grounded within a shared in-distribution learning loop.

Agentic Large Language Model (LLM) systems decompose complex tasks into workflow Directed Acyclic Graphs (DAGs) whose primitives must be scheduled on heterogeneous clusters. Existing deep reinforcement learning (DRL) schedulers are tied to a fixed cluster size and require retraining whenever the number of servers changes. We propose SCALE (Scalable Cross-Attention Learning with Extrapolation), a DRL scheduler that generalizes to unseen cluster scales without fine-tuning. SCALE employs a cross-attention pointer network where task features query against server features, so the architecture accepts any number of servers by construction. We observe, however, that permutation-invariant architecture alone does not guarantee good performance at new scales - the attention feature undergoes distribution shift as the server count grows. To counter this, we introduce Structured Representation Regularization (SRR): a decorrelation loss combined with a KL penalty toward the standard normal, which keeps feature statistics stable regardless of input size. Trained on 16 nodes and tested directly on 32 and 48 nodes, SCALE reduces average response time by 8.9% at N=48 relative to the same architecture without SRR, confirming that explicit regularization is necessary to close the scale-generalization gap.

Long-horizon LLM agents require reinforcement learning methods that can assign credit to intermediate decisions under sparse and delayed rewards. Recent group-based methods such as GiGPO improve over GRPO by constructing step-level advantages at repeated anchor states. However, we show that such dense credit can be statistically unreliable: under limited rollouts, rare but lucky actions may receive overly large advantages, producing divergent anchor bias and late-stage training oscillation. We propose Evidence-Calibrated Policy Optimization (ECPO), a critic-free policy optimization algorithm that calibrates step-level credit before policy updates. ECPO combines Evidence-Calibrated Action Advantage, which groups rollouts by canonical actions and shrinks low-count estimates, with Variance-Gated Credit Weighting, which suppresses anchor states dominated by within-action noise. Experiments on ALFWorld and WebShop with Qwen2.5-1.5B/7B show that ECPO consistently outperforms strong baselines, improving GiGPO by +5.2/+7.3 success points on ALFWorld/WebShop with Qwen2.5-1.5B while adding only 0.1% additional advantage-computation overhead.

LLM agents operate in two distinct regimes: open-weight agents amenable to reinforcement learning (RL) and black-box agents whose behaviour must be controlled purely at test time. Although black-box agents are often backed by state-of-the-art proprietary LLMs, API-only access precludes parameter-level optimization, rendering most RL methods inapplicable. To address this limitation, we turn to a known equivalence between RL and Bayesian inference. We propose Agentic Monte Carlo (AMC) to directly sample from the optimal policy of a black-box agent rather than training it through RL. The optimal policy is a posterior over trajectories whose prior we define as the fixed black-box LLM agent. We employ Sequential Monte Carlo to sample from this posterior by learning a value function to steer the agent while leaving the underlying black-box model unchanged. We validate AMC on three diverse environments from the AgentGym benchmark, demonstrating significant improvements over prompting baselines and even outperforming Group Relative Policy Optimization (GRPO) as we scale the test-time compute of our method. AMC demonstrates the feasibility of performing principled RL-style optimization of black-box LLM agents. Code is available at https://github.com/layer6ai-labs/Agentic-Monte-Carlo

Reinforcement learning with verifiable rewards improves reasoning and tool use, yet long-horizon language agents still learn unsupported evidence chains, belief drift, and shortcut actions that satisfy terminal checks. Existing process rewards are mostly correlational: they reward retrieval-, reflection-, or verification-like steps without estimating whether the step contributes to final verified success under a specified intervention. We propose CVT-RL, a constrained policy-gradient algorithm with dense verifiable rewards, intervention-validity gating, and a policy-conditioned counterfactual contribution (PCCC) estimator. Deletion, semantic substitution, evidence substitution, and tool-output perturbation define separate controlled interventions; continuations are sampled from a frozen reference policy, and a selection-adjusted doubly robust estimator augments the advantage. Belief control uses only prefix-observable labels, while an augmented Lagrangian constrains unsupported claims, skipped verification, tool tampering, and unsafe calls. On long-context QA, ALFWorld, ScienceWorld, and web/tool tasks, CVT-RL improves average task success from 71.8% for compute-matched non-causal RL and 75.4% for an information-matched counterfactual-process baseline to 78.9%, improves evidence F1 from 78.9 to 82.8 over the information-matched baseline, and reduces measured hacking from 7.2% to 3.9%. Independent human audit estimates 4.6% hacking for CVT-RL versus 8.1% for the information-matched baseline, and adaptive detector-evasion attacks raise hacking only to 7.1%. Stratified bootstrap and mixed-effects tests give p<0.01 after Holm correction for all primary metrics. Carefully scoped counterfactual credit, paired with validity gating, diagnostics, and verifiable constraints, provides a reproducible route toward more reliable long-horizon RL for language agents.

Lifelong learning is essential for Large Language Model (LLM) agents operating in dynamic, interactive environments. However, existing lifelong learning agents for long-horizon tasks typically depend on discrete skill or past experiences retrieval with static parameters during inference, which prevents them from continuously internalizing test-time feedback like human learners. To bridge this gap, we propose Skill-enhanced Test-Time Co-Evolution (\texttt{LifeSkill}), a two-stage reinforcement learning framework for Online Lifelong Learning Agents. Specifically, we design Verifier-Guided Skill Learning that addresses the lack of direct supervision for skill extraction by rewarding candidate skills according to the average verifier success of multiple skill-conditioned policy rollouts, encouraging the model to generate skills that are useful for solving tasks rather than merely plausible in text. Furthermore, we introduce Online Skill Internalization, which continuously improves the policy model during test-time interaction by transforming skill-conditioned trajectories into reward signals. This enables the agent to directly internalize reasoning capabilities into its parameters, avoiding the context bloat of experience retrieval. Experiments on LifelongAgentBench show that LifeSkill improves average performance by 7 absolute points by comparing with existing lifelong agent baselines.

Existing memory-augmented LLM agents store past experience exclusively in prompt space, as textual summaries or retrieved passages, while keeping model parameters frozen throughout a rollout. Such agents can \emph{look up} what they have seen but cannot \emph{learn from} it: their policy is unchanged by experience, and any information dropped from the context is permanently lost. We introduce \texttt{TMEM}, a self-evolving parametric memory framework in which the agent not only compresses history into explicit memory but also absorbs distilled supervision into fast LoRA weights $Δ_t$ via lightweight online updates, genuinely altering its future behavior within a single episode. We formalize this as an agentic decision process with fast-weight rollout dynamics: actions are sampled from $π_{θ_0+Δ_t}$, while extraction actions produce supervision that updates $Δ_t$ for subsequent decisions. This view makes the extraction policy directly optimizable by RL: training $θ_0$ improves not only task actions but also the quality of the data used for online LoRA adaptation. We further propose SVD-based initialization of the LoRA subspace to accelerate online convergence. Experiments on LoCoMo, LongMemEval-S, multi-objective search, and CL-Bench show that \texttt{TMEM} consistently outperforms summary-based and retrieval-based baselines across different model scales.

The supply of high-quality training tasks is a central bottleneck for reinforcement learning from verifiable rewards (RLVR) on agentic language models. Each task requires a sandboxed setup, a prompt, and a hand-authored reward function, and only tasks that pass a quality bar produce useful training signal. Hand-curation at this quality bar does not scale economically to the task counts effective RL training requires, and the substitution rate between automatically generated task variants and human-authored ones is not yet established. We investigate using pre-specified, gate-filtered augmentations of a small hand-authored base as a substitute for additional human curation during RLVR. We formalize the cost-adjusted trade rate $ρ_{\text{cost}}$ between augmented and human-authored tasks, measure it through a controlled ablation across training corpora with varying augmentation share, and characterize the end-to-end economics of the augmentation pipeline. Substituting augmented content for additional human-authored tasks retains aggregate held-out generalization on a ten-benchmark suite spanning code, instruction following, reasoning, and multi-turn agentic function-calling. The cost-adjusted trade rate $ρ_{\text{cost}}$ between gated synthetic and human-authored RLVR tasks stays in $[1.4\times, 11.6\times]$ across the plausible $c_{\text{human}}/c_{\text{aug}}$ range.

A central goal of large language model (LLM) research is to build agentic systems that can plan, act, and adapt through sustained interaction with dynamic environments. While recent LLM-based agents exhibit impressive contextual reasoning, their long-horizon decision-making remains fragile, often suffering from objective drift, where goals and plans drift over extended interactions. We introduce Multi$^2$, a hierarchical multi-agent decision-making framework that explicitly decomposes agent behavior into complementary roles. A high-level agent (System 1) focuses on context-aware sub-goal generation using supervised fine-tuning (SFT), while a low-level agent (System 2) executes atomic actions through offline-to-online reinforcement learning (RL) in interactive environments. This separation enables stable long-horizon control, mitigates objective drift, and allows efficient adaptation. Across diverse interactive environments, Multi$^2$ consistently outperforms strong agentic baselines, demonstrating improved robustness and coordination in multi-turn interaction. Beyond performance, we introduce and release three hierarchical benchmark datasets, filling a long-standing gap in training and evaluating hierarchical decision-making for LLM-based agents.

Long-context tasks require LLMs to identify and preserve answer-relevant information from large contexts. Chunk-wise memory agents address this issue by sequentially reading document chunks, updating a compact memory, and generating the final answer from the accumulated memory. However, existing RL-based chunk-wise agents either rely on sparse final-answer rewards or use lexical intermediate rewards for memory and retrieval actions. These signals supervise task success or local overlap, but do not directly evaluate whether the final memory supports the ground-truth answer. We propose InfoMem, a reward mechanism for training chunk-wise memory agents that evaluates final-memory utility using answer-conditioned information. InfoMem measures how much the final memory increases the model's per-token log-likelihood of the ground-truth answer. To stabilize RL optimization, InfoMem applies this signal only to successful trajectories and normalizes it before reward composition. Under the same GRPO framework and training budget, InfoMem improves long-context memory-agent performance over comparable memory-agent RL baselines. Analyses show that effective final-memory rewards should operate on successful trajectories, be normalized before reward composition, and be conditioned on the answer rather than the query. Our code is available at https://github.com/GenSouKa1/InfoMem.

LLM-based search agents are trained predominantly with outcome-only reward, leaving the search process itself unsupervised. This signal degenerates on outcome-homogeneous groups where all sampled trajectories share the same correctness, yielding zero within-group advantage and no gradient. Existing process supervision either trains a costly verifier or generates per-query rubrics that are inconsistent across queries and discarded after one use. We propose ARBOR (Adaptive Rubric Buffer for Online Reward), a reusable process-reward framework that maintains a rubric memory shared across queries. Query-local drafts induced from contrastive trajectories are admitted, consolidated into cross-query common rubrics, and retired as the policy evolves. A small active subset of common rubrics scores trajectories via sparse pairwise judging, and the resulting scores are added to the base reward, providing process-level gradient even when outcome reward is uniform. ARBOR consistently outperforms GRPO and DAPO baselines on four multi-hop QA benchmarks, raising average LLM-judge accuracy by up to 4.2 points and converting up to 42% of otherwise-zero-gradient training groups into informative ones.

Reinforcement learning (RL) improves large language model (LLM) agents by teaching them which actions lead to high rewards, but provides little supervision on what those actions do to the environment. World modeling (WM) can fill this gap, yet existing approaches often require separate simulators, extra training stages, or additional inference-time computation. We observe that on-policy RL rollouts already contain the needed signal: each transition pairs an action with its resulting next observation. Based on this observation, we propose PaW, a Policy and World modeling co-training framework that adds auxiliary WM supervision to the same policy during RL, without changing the inference paradigm. To make auxiliary WM supervision informative and stable, PaW introduces three components: action-entropy-based WM data selection, noise-tolerant WM loss, and reward-adaptive loss balancing. Experiments on three agentic task benchmarks show consistent improvements over strong RL baselines across models and RL algorithms. These results suggest that standard RL rollouts are a practical source of WM supervision for language-agent training.

Search agents are often trained as policies over growing transcripts: the model must decide how to search while also remembering what it has seen, which evidence is useful, which constraints remain open, and which claims have actually been checked. We argue that this formulation puts too much routine state management inside the policy: reinforcement learning is forced to optimize both semantic search decisions and recoverable bookkeeping that the environment can maintain more reliably. We introduce Harness-1, a 20B search agent (retrieval subagent) trained with reinforcement learning inside a stateful search harness. The harness maintains environment-side working memory, including a candidate pool, an importance-tagged curated set, compact evidence links, verification records, compressed and deduplicated observations, and budget-aware context rendering. The policy retains the semantic decisions: what to search, which documents to keep or discard, what to verify, and when to stop. Across eight retrieval benchmarks spanning web, finance, patents, and multi-hop QA, Harness-1 achieves 0.730 average curated recall, outperforming the next strongest open search subagent by +11.4 points and remaining competitive with much larger frontier-model searchers. Its gains are especially strong on held-out transfer benchmarks, suggesting that reinforcement learning over explicit search state can produce retrieval behaviors that generalize beyond the training domains. Our code is available at https://github.com/pat-jj/harness-1.

Long-horizon LLM agents can benefit from reusable skills, yet existing skill-based methods often rely on external skill generators during training or persistent skill retrieval at inference, increasing engineering complexity, context length, and deployment latency. We propose Self-Internalizing Reinforcement learning with Intrinsic skills (SIRI), a three-phase framework that enables agents to discover, validate, and internalize skills without external skill generators or inference-time skill banks. SIRI first warms up the policy with GiGPO to acquire basic interaction ability and collect successful skill-free trajectories. It then performs self-skill mining, where the current policy summarizes compact skills from its own successful plain rollouts and validates them through paired skill-augmented and skill-free rollouts. Finally, SIRI distills only beneficial skill-guided action tokens into the plain policy using trajectory-level utility and action-level advantage. At inference, the agent runs with the original prompt only. On ALFWorld and WebShop with Qwen2.5-7B-Instruct, SIRI improves GiGPO from 0.908 to 0.930 on ALFWorld and from 0.728 to 0.813 on WebShop, outperforming prompt-based, RL-based, and memory-augmented baselines. Further analysis shows that our self-mining strategy can achieve performance comparable to distillation with closed-source large model. Our code is available at https://github.com/kirito618/SIRI.

Agentic reinforcement learning (RL) enables LLM agents to improve continuously from environment rewards, yet the resulting policies do not systematically accumulate reusable strategies that generalize across tasks. Modular skills can provide such reusable strategies, yet existing skill-augmented RL methods decouple skill creation from policy optimization, risking adopting skills that conflict with the evolving policy. Inspired by Anthropic's Skill Creator, we introduce ReSkill, an RL-in-the-loop skill creation framework that reconciles skill evolution with policy learning. ReSkill exploits the group-wise structure of GRPO to naturally embed three mechanisms with only marginal additional overhead: (1) an assertion-driven skill creator that diagnoses failures from past experience and proposes conditional, trigger-based skill revisions; (2) within-group rollout sampling that enables controlled comparison of skill versions, capturing which version best supports the policy's ongoing learning; and (3) Thompson Sampling with adaptive discounting to balance exploration and exploitation in skill version selection as the policy evolves. Across several domains, ReSkill consistently outperforms existing memory and skill-based RL methods, with the largest gains on unseen tasks. Analysis of the skill lifecycle shows skills being automatically created, tested, refined, and pruned as the policy improves, demonstrating reconciled skill-policy co-evolution.

Large language model agents are moving beyond text-only interaction toward physical-world control, with smart homes as a representative domain. Real domestic interaction requires understanding ambiguous intents, operating in dynamic environments, and performing multi-turn reasoning. However, existing methods struggle to generate high-quality training data for smart home agents. We propose HomeFlow, a verifiable data flywheel for this domain. HomeFlow uses HomeEnv as a unified simulation environment and HomeMaker to procedurally generate diverse home settings. Subsequently, Blueprint compiles open-ended user intents into executable state-based success conditions, while MCTS-Flow synthesizes diverse, verifiable multi-turn trajectories through environment-guided tree search. We then optimize the agents via supervised fine-tuning and step-wise RLVE, which facilitates iterative improvement through authentic physical feedback. We further construct SmartHome-Bench to evaluate the agent across various smart home tasks. On this benchmark, HomeFlow-RL-4B and HomeFlow-RL-8B achieve task success rates of 84.60% and 87.03%. It is worth noting that HomeFlow-RL-8B even surpasses the leading GPT-5.5 by 1.23 percentage points.

Interactive text-to-SQL agents solve database tasks through multi-turn interactions involving schema exploration, query execution, feedback interpretation, and decision revision. Long-term memory helps agents reuse past experiences, but existing retrieval methods remain limited. Static methods rely on fixed similarity heuristics that do not optimize downstream utility, while dynamic methods often learn from sparse final outcomes and retrieve memories at a single decision horizon. This is insufficient when memory usefulness changes across interaction stages, since memories useful for initial planning may differ from those needed for local, state-conditioned execution. We propose MERIT, a dynamic multi-horizon memory retrieval framework. MERIT maintains episode-level memory for global strategic guidance and turn-level memory for local decision support. Both levels use learned retrieval policies optimized with reinforcement learning. To train turn-level retrieval despite limited intermediate supervision, MERIT uses a lightweight Process Reward Model to provide dense proxy rewards for local memory selection. Experiments on BIRD-Interact show that MERIT outperforms no-memory, static-retrieval, and dynamic-retrieval baselines in success rate while reducing average interaction turns. Transfer results on Spider2-Snow further show positive cross-benchmark transfer without benchmark-specific tuning. These results suggest that multi-horizon retrieval improves experience reuse in interactive text-to-SQL agents.

Large language model agents trained with reinforcement learning (RL) often learn brittle, task-specific shortcuts. We hypothesize that agents generalize better when their successful trajectories are structurally compressible, decomposed into a small set of reusable abstract patterns. To formalize this, we introduce ReuseRL, which grounds agentic RL in the Minimum Description Length (MDL) principle. ReuseRL extracts a shared skill dictionary from successful trajectories and augments the RL objective with a segmentation cost, explicitly penalizing idiosyncratic behaviors that encode poorly. We prove a PAC-Bayes bound guaranteeing that a dictionary extracted from successful trajectories has bounded expected description length on future successful behavior. Across ALFWorld, TextWorld-Cooking, and Countdown-Stepwise, ReuseRL improves in- and out-of-distribution success over vanilla GRPO and strong round-length baselines.

Long-term memory is essential for multimodal agents to build coherent experience, accumulate world knowledge, and achieve continual learning. However, constructing effective memory goes beyond memory module design and basic requirements such as accuracy and fidelity; the key challenge lies in determining what to memorize. Multimodal agents, such as embodied agents, continuously perceive, reason, and act in real or virtual environments, receiving an unbounded stream of multimodal observations. From this combinatorial explosion of information, an agent must selectively retain content that is relevant to its role in the environment and valuable for future tasks. To bridge this gap, we frame memory generation as a learnable memorization policy and introduce TaskMem (Task-focused Memorization Policy Learning), a reinforcement-learning-based framework that enables the policy to dynamically adjust its focus to the demands of real tasks encountered in the environment. TaskMem adopts a two-phase training paradigm: Phase One learns how to memorize by optimizing memory quality under fundamental fidelity requirements; Phase Two occurs after deployment, where the agent learns what to memorize by tuning an adapter on its base MLLM, using recent environment tasks to define a reward model that guides the memorization policy toward task-relevant content. To evaluate our approach, we reformulate VideoMME, EgoLife, and EgoTempo into streaming benchmarks that simulate a realistic setting in which an agent processes streaming observations and handles tasks arriving online. To isolate memory assessment, the questions must be answered using only the agent's memory, without access to raw video. Built on Qwen3-VL-30B-A3B, TaskMem improves VQA accuracy by 6.3%, 7.0%, and 5.3% on these benchmarks, respectively.