Long-Horizon Agent Tasks

Recent GUI agents have made substantial progress in visual grounding and action prediction, yet they remain brittle in long-horizon tasks that require maintaining task state across many interface transitions. Existing agents typically rely on raw history replay or text-only memory, which either overwhelms the model with redundant screenshots or discards localized visual evidence needed for future decisions. To address these limitations, we introduce MementoGUI, a plug-in agentic memory framework that equips MLLM-based GUI agents with MementoCore, a learned controller for online memory selection, compression, and retrieval. Rather than treating interaction history as a fixed context, MementoGUI formulates long-horizon GUI control as an online memory-control problem: working memory selectively preserves task-relevant interface events with textual summaries and ROI-level visual evidence, while episodic memory retrieves reusable past trajectories through learned relevance selection. MementoCore modularizes memory control into specialized operators for step processing, memory compression, episodic writing, and episodic selection, enabling plug-in memory augmentation without finetuning the GUI agent backbone. We further develop a scalable data curation pipeline that converts computer-use trajectories into memory-controller training data, introduce MementoGUI-Bench for evaluating long-horizon decision-making in GUI agents, and design MLLM-based metrics for semantic action matching, task progress, and memory consistency. Experiments on GUI-Odyssey, MM-Mind2Web, and MementoGUI-Bench show that MementoGUI consistently improves GUI agents over no-history, history-replay, and text-only memory baselines, with larger MementoCore backbones further strengthening memory-augmented GUI control.

Long-horizon multimodal agents in open-world games must stay goal-directed across many low-level interactions under tight token and latency budgets. Existing approaches often trade off costly per-step reasoning against reactive execution that can drift, repeat failures, and recover poorly. Our key idea is to reuse strategic reasoning across locally stable segments and reinvoke it at event boundaries. We present SPIKE, an adaptive dual controller framework for cost-efficient long-horizon game control. Its Strategic Controller performs low-frequency global planning, failure analysis, and recovery, while its Reactive Controller handles fast local execution under a strict token budget. An Event Trigger monitors visual change, task progress, repeated actions, and failure signals to decide when control should stay reactive or escalate to strategic reasoning. Hierarchical Memory separates short-term experience reuse in the State-Action Memory Bank (SA-MB) from structured evidence in the State Action Knowledge Graph (SA-KG), allowing each controller to retrieve the context it needs. This design reuses strategic proposals over multiple reactive steps, supports local override when plans become stale, and reserves expensive reasoning for moments where extra deliberation is useful. On the Lite-100 split of StarDojo, SPIKE improves Lite-100 success rate (SR) by 5.0 percentage points (38.5% relative) over the strongest Lite-100 baseline and Budgeted SR by 9.3 points (75.6% relative) over the strongest budgeted baseline. It also reduces token consumption by 54.9% and latency by 40.8%. Ablations show that event triggering, reactive override, and heterogeneous memory each contribute to success and recovery, supporting selective reasoning rather than reasoning at every step.

Coding agents are increasingly deployed in real software development, where a single version iteration requires months of coordinated work across many files. However, most existing benchmarks focus predominantly on single-issue bug fixes from Python repositories, with coarse pass/fail evaluation outcomes, and thus fail to capture long-horizon, multi-target development at real engineering scale. To address this gap, we present RoadmapBench, a benchmark of 115 long-horizon coding tasks grounded in real open-source version upgrades across 17 repositories and 5 programming languages. Each task places the agent on a source-version code snapshot and provides a multi-target roadmap instruction requiring it to implement the functionality introduced in the target version, with a median modification of 3,700 lines across 51 files. We conduct a systematic evaluation on thirteen frontier models and find that even the strongest, Claude-Opus-4.7, resolves only 39.1% of tasks, while the weakest achieves merely 5.2%, in stark contrast to existing bug-fix benchmarks, suggesting that long-horizon software development remains a largely unsolved problem.

Long-horizon visuomotor tasks remain a fundamental challenge in AI, as current methods rely on hand-engineered rewards or action-labeled demonstrations, neither of which scales. We introduce ASH, an agentic system that learns a long-horizon policy from unlabeled, noisy internet video, without reward shaping or expert annotation. ASH follows a self-improvement loop; when it gets stuck, ASH learns an Inverse Dynamics Model (IDM) from its own trajectories, and uses its IDM to extract supervision from relevant internet video. ASH uses unsupervised learning to identify key moments from large-scale internet video and retains them as long-term memory - allowing it to tackle long-horizon problems. We evaluate ASH on two complementary environments demanding multi-hour planning: Pokemon Emerald, a turn-based RPG, and The Legend of Zelda: The Minish Cap, a real-time action-adventure game. In both games, behavioral cloning, retrieval-augmented and zero-shot foundation-model baselines plateau, while ASH sustains progression across our 8-hour evaluation. ASH reaches an average of 11.2/12 milestones in Pokemon Emerald and 9.9/12 in Legend of Zelda, while the strongest baseline gets stuck in both environments at an average of 9.3/12 and 7.8/12 milestones, respectively. We demonstrate that self-improving agents are a scalable recipe for learning in long-horizon worlds.

Agentic evolution has emerged as a powerful paradigm for improving programs, workflows, and scientific solutions by iteratively generating candidates, evaluating them, and using feedback to guide future search. However, existing methods are typically instantiated either as fixed hand-designed procedures that are modular but rigid, or as general-purpose agents that flexibly integrate feedback but can drift in long-horizon evolution. Both forms accumulate rich evidence over time, including candidates, feedback, traces, and failures, yet lack a stable interface for organizing this evidence and revising the mechanism that drives future evolution. We address this limitation by formulating agentic evolution as an interactive environment, where the accumulated evolution context serves as a process-level state. We introduce AEvo, a harnessed meta-editing framework in which a meta-agent observes this state and acts not by directly proposing the next candidate, but by editing the procedure or agent context that controls future evolution. This unified interface enables AEvo to steer both procedure-based and agent-based evolution, making accumulated evidence actionable for long-horizon search. Empirical evaluations on agentic and reasoning benchmarks show that AEvo outperforms five evolution baselines, achieving a 26 relative improvement over the strongest baseline. Across three open-ended optimization tasks, AEvo further outperforms four evolution baselines and achieves state-of-the-art performance under the same iteration budget.

Computer-use agents (CUAs) rely on visual observations of graphical user interfaces, where each screenshot is encoded into a large number of visual tokens. As interaction trajectories grow, the token cost increases rapidly, limiting the amount of history that can be incorporated under fixed context and compute budgets. This has resulted in no or very limited improvement in the performance when using history unlike other domains. We address this inefficiency by introducing ReVision, which is used to train multimodal language models on trajectories where redundant visual patches are removed using a learned patch selector that compares patch representations across consecutive screenshots while preserving spatial structure required by the model. Across three benchmarks, OSWorld, WebTailBench, and AgentNetBench, when processing trajectories with 5 history screenshots using Qwen2.5-VL-7B, ReVision reduces token usage by 46% on average while improving success rate by 3% over the no drop baseline. This establishes a clear efficiency gain, enabling agents to process longer trajectories with fewer tokens. With this improved efficiency, we revisit the role of history in CUAs and find that performance continues to improve as more past observations are incorporated when redundancy is removed.

Imitation learning is effective for training agents when expert demonstrations are available, but collecting demonstrations for every complex task in an environment is costly. We study the long-horizon, goal-conditioned setting where a fixed demonstration dataset contains useful behavior, but not complete examples for every task the agent must solve. Existing imitation learning methods can learn strong policies from demonstrations, but when solving long-horizon tasks, small errors accumulate over long primitive-action trajectories and make zero-shot adaptation to new tasks unreliable. We introduce Zero-shot Agents from Latent Topologies (ZALT), an imitation-learning method that solves unseen start-goal tasks beyond those demonstrated during training. ZALT identifies latent hub states where trajectories converge or diverge, learns policies and a dynamics model over hub-to-hub transitions, and plans over the hub topology to complete new tasks. This topology makes demonstrated behaviors explicitly composable while compressing long tasks into shorter sequences of abstract transitions -- combined, these enable ZALT to perform zero-shot adaptation. In a complex 3D maze environment, ZALT achieves 55% zero-shot success on unseen tasks, compared to 6% for the strongest baseline.

Browser agents are increasingly deployed in long-horizon tasks, which require executing extended action chains to accomplish user goals. However, this prolonged execution process provides attackers with more opportunities to inject malicious instructions. Existing prompt injection attacks against browser agents expose two key gaps: (1) low effectiveness, as attacks optimized for toy baselines fail to achieve end-to-end goals in real-world scenarios with complex environments and longer steps; (2) weak stealthiness, since most attacks pit the attack goal against the user goal, causing a significant drop in system usability under attack. To address these gaps, we propose WebTrap, a mid-task hijacking injection attack. It employs multi-step instruction fusion steering to seamlessly combine both goals, enabling the agent to resume the original user task after executing the attack goal. Furthermore, we design a context-grounded generation method to align the injected content with the task environment and system instructions, maximizing the hijacking success rate. Extensive experiments on two browser agent tasks, based on extended WASP and InjecAgent environments, demonstrate that our method achieves a high attack success rate while preserving the usability of the original system. We find that WebTrap exploits the agent's navigation vulnerabilities, binding the two goals so tightly that standard defense mechanisms cannot restore the system to normal operation. These findings reveal a critical vulnerability in agent systems during long-horizon tasks that they can be stealthily hijacked.

While long-horizon agentic tasks require language agents to perform dozens of sequential decisions, training such agents with reinforcement learning remains challenging. We identify two root causes: credit misattribution, where correct early actions are penalized due to terminal failures, and sample inefficiency, where scarce successful trajectories result in near-total loss of learning signal. We introduce a milestone-guided policy learning framework, BEACON, that leverages the compositional structure of long-horizon tasks to ensure precise credit assignment. BEACON partitions trajectories at milestone boundaries, applies temporal reward shaping within segments to credit partial progress, and estimates advantages at dual scales to prevent distant failures from corrupting the evaluation of local actions. On ALFWorld, WebShop, and ScienceWorld, BEACON consistently outperforms GRPO and GiGPO. Notably, on long-horizon ALFWorld tasks, BEACON achieves 92.9% success rate, nearly doubling GRPO's 53.5%, while improving effective sample utilization from 23.7% to 82.0%. These results establish milestone-anchored credit assignment as an effective paradigm for training long-horizon language agents. Code is available at https://github.com/ZJU-REAL/BEACON.

Large language models (LLMs) have shown promise as interactive agents that solve tasks through extended sequences of environment interactions. While prior work has primarily focused on system-level optimizations or algorithmic improvements, the role of task horizon length in shaping training dynamics remains poorly understood. In this work, we present a systematic empirical study that examines horizon length through controlled task constructions. Specifically, we construct controlled tasks in which agents face identical decision rules and reasoning structures, but differ only in the length of action sequences required for successful completion. Our results reveal that increasing horizon length alone constitutes a training bottleneck, inducing severe training instability driven by exploration difficulties and credit assignment challenges. We demonstrate that horizon reduction is a key principle to address this limitation, stabilizing training and achieving better performance in long-horizon tasks. Moreover, we find that horizon reduction is related to stronger generalization across horizon lengths: models trained under reduced horizons generalize more effectively to longer-horizon variants at inference time, a phenomenon we refer to as horizon generalization.

Foundation model-driven agents often struggle with long-horizon planning due to the transient nature of purely prompting-based reasoning. While existing skill induction methods mitigate this by distilling experience into state-blind parameterized scripts, they fail to capture the conditional logic required for robust execution in dynamic environments. In this paper, we propose Neuro-Symbolic Skill Induction (NSI), a framework that lifts interaction traces into modular, logic-grounded programs. By synthesizing explicit control flows and dynamic variable binding, NSI empowers agents to discover when and why to act. This paradigm enables the efficient generalization, allowing agents to induce skills from few-shot examples and flexibly adapt to unseen goals. Experiments on a series of agentic tasks demonstrate that NSI consistently outperforms state-of-the-art baselines, empowering agents to self-evolve into architects of logic-grounded skills.

Realistic long-horizon productivity work is strongly conditioned on user-specific computer environments, where much of the work context is stored and organized through directory structures and content-rich artifacts. To scale synthetic data creation for such productivity scenarios, we introduce Synthetic Computers at Scale, a scalable methodology for creating such environments with realistic folder hierarchies and content-rich artifacts (e.g., documents, spreadsheets, and presentations). Conditioned on each synthetic computer, we run long-horizon simulations: one agent creates productivity objectives that are specific to the computer's user and require multiple professional deliverables and about a month of human work; another agent then acts as that user and keeps working across the computer -- for example, navigating the filesystem for grounding, coordinating with simulated collaborators, and producing professional artifacts -- until these objectives are completed. In preliminary experiments, we create 1,000 synthetic computers and run long-horizon simulations on them; each run requires over 8 hours of agent runtime and spans more than 2,000 turns on average. These simulations produce rich experiential learning signals, whose effectiveness is validated by significant improvements in agent performance on both in-domain and out-of-domain productivity evaluations. Given that personas are abundant at billion scale, this methodology can in principle scale to millions or even billions of synthetic user worlds with sufficient compute, enabling broader coverage of diverse professions, roles, contexts, environments, and productivity needs. We argue that scalable synthetic computer creation, together with at-scale simulations, is highly promising as a foundational substrate for agent self-improvement and agentic reinforcement learning in long-horizon productivity scenarios.

Computer-use agents provide a promising path toward general software automation because they can interact directly with arbitrary graphical user interfaces instead of relying on brittle, application-specific integrations. Despite recent advances in benchmark performance, strong computer-use agents remain expensive and slow in practice, since most systems invoke large multimodal models at nearly every interaction step. We argue that this uniform allocation of compute is fundamentally inefficient for long-horizon GUI tasks. Such trajectories are highly heterogeneous: many steps are routine and can be handled reliably by smaller, cheaper policies, while errors tend to concentrate at a relatively small number of high-risk moments. Across computer-use benchmarks, these failures repeatedly take two forms: progress stalls, where the agent loops, repeats ineffective actions, or fails to make meaningful progress, and silent semantic drift, where the agent continues taking locally plausible actions after already deviating from the user's true goal. To address this inefficiency, we propose an event-driven, step-level cascade for computer-use agents that runs a small policy by default and escalates to a stronger model only when lightweight learned monitors detect elevated risk. Our framework combines two complementary signals: a Stuck Monitor that detects degraded progress from recent reasoning-action history and triggers recovery, and a Milestone Monitor that identifies semantically meaningful checkpoints where sparse verification is most informative for catching drift. This design turns always-on frontier-model inference into adaptive, on-demand compute allocation over the course of an evolving interaction. The framework is modular and deployment-oriented: it can be layered on top of existing computer-use agents without changing the underlying agent architecture or retraining the large model.

Long-horizon LLM tasks often fail not because a single answer is unattainable, but because knowledge states drift across rounds, intermediate commitments remain implicit, and interruption fractures the evolving evidence chain. This paper presents ADEMA as a knowledge-state orchestration architecture for long-horizon knowledge synthesis rather than as a generic multi-agent runtime. The architecture combines explicit epistemic bookkeeping, heterogeneous dual-evaluator governance, adaptive task-mode switching, reputation-shaped resource allocation, checkpoint-resumable persistence, segment-level memory condensation, artifact-first assembly, and final-validity checking with safe fallback. Evidence is drawn entirely from existing materials: a four-scenario showcase package, a fixed 60-run mechanism matrix, targeted micro-ablation and artifact-chain supplements, and a repaired protocol-level benchmark in which code-oriented evaluation is the clearest quality-sensitive mechanism block. Across the fixed matrix, removing checkpoint/resume produced the only invalid run, and it did so in the interruption-sensitive resume condition. By contrast, dual evaluation, segment synthesis, and dynamic governance are best interpreted as supporting control mechanisms that shape trajectory discipline, explicit artifact progression, and cost-quality behavior rather than as universal binary prerequisites for completion. The contribution is therefore a knowledge-state orchestration architecture in which explicit epistemic state transition, evidence-bearing artifact progression, and recoverable continuity are the primary design commitments.

Existing web agent benchmarks have largely converged on short, single-site tasks that frontier models are approaching saturation on. However, real world web use consists of long-horizon, multi-site workflows. Common web navigation tasks, such as comparing products across different domains, planning trips across multiple services, or summarizing information from multiple search queries, require sustained context and cross-site reasoning over potentially hours of browsing. To capture and evaluate such behaviors, we introduce Odysseys: a benchmark of 200 long-horizon web tasks derived from real world browsing sessions evaluated on the live Internet. We find that binary pass/fail evaluation is inadequate for long-horizon settings and introduce a rubric-based evaluation, annotating each Odysseys task with an average of 6.1 graded rubrics. We demonstrate that this yields higher agreement with humans and provides a more fine-grained signal than commonly used trajectory-level LLM-as-a-judge evaluation metrics. We tested several leading frontier models and find that the strongest models achieve a success rate of 44.5%, which leaves substantial room for future improvements. Beyond task success, we argue that efficiency is a first-class concern for long-horizon agents. We introduce a Trajectory Efficiency metric (rubric score per step) and find that even frontier agents achieve only 1.15%, marking an evident need for agents that can succeed efficiently and not simply eventually. Odysseys isolates the critical evaluation of long-horizon proficiency in open-web environments, providing a realistic benchmark to measure progress towards computer-use agents that can potentially productively operate for hours. We release our tasks, evaluation scripts, and other results at https://odysseys-website.pages.dev

Large language model-based agents have recently emerged as powerful approaches for solving dynamic and multi-step tasks. Most existing agents employ planning mechanisms to guide long-term actions in dynamic environments. However, current planning approaches face a fundamental limitation that they operate at a fixed granularity level. Specifically, they either provide excessive detail for simple tasks or insufficient detail for complex ones, failing to achieve an optimal balance between simplicity and complexity. Drawing inspiration from the principle of progressive refinement in cognitive science, we propose AdaPlan-H, a self-adaptive hierarchical planning mechanism that mimics human planning strategies. Our method initiates with a coarse-grained macro plan and progressively refines it based on task complexity. It generates self-adaptive hierarchical plans tailored to the varying difficulty levels of different tasks, which can be optimized by imitation learning and capability enhancement. Experimental results demonstrate that our method significantly improves task execution success rates while mitigating overplanning at the planning level, providing a flexible and efficient solution for multi-step complex decision-making tasks. To contribute to the community, our code and data will be made publicly available at https://github.com/import-myself/AHP.

As Multimodal Large Language Models (MLLMs) mature, GUI agents are evolving from static interactions to complex navigation. While Reinforcement Learning (RL) has emerged as a promising paradigm for training MLLM agents on dynamic GUI tasks, its effective application faces a dilemma. Standard Offline RL often relies on static step-level data, neglecting global trajectory semantics such as task completion and execution quality. Conversely, Online RL captures the long-term dynamics but suffers from high interaction costs and potential environmental instability. To bridge this gap, we propose SOLAR-RL (Semi-Online Long-horizon Assignment Reinforcement Learning). Instead of relying solely on expensive online interactions, our framework integrates global trajectory insights directly into the offline learning process. Specifically, we reconstruct diverse rollout candidates from static data, detect the first failure point using per-step validity signals, and retroactively assign dense step-level rewards with target-aligned shaping to reflect trajectory-level execution quality, effectively simulating online feedback without interaction costs. Extensive experiments demonstrate that SOLAR-RL significantly improves long-horizon task completion rates and robustness compared to strong baselines, offering a sample-efficient solution for autonomous GUI navigation.

Long-Horizon (LH) tasks in Human-Scene Interaction (HSI) are complex multi-step tasks that require continuous planning, sequential decision-making, and extended execution across domains to achieve the final goal. However, existing methods heavily rely on skill chaining by concatenating pre-trained subtasks, with environment observations and self-state tightly coupled, lacking the ability to generalize to new combinations of environments and skills, failing to complete various LH tasks across domains. To solve this problem, this paper presents ALAS, a cross-domain learning framework for LH tasks via biologically inspired dual-stream disentanglement. Inspired by the brain's "where-what" dual pathway mechanism, ALAS comprises two core modules: i) an environment learning module for spatial understanding, which captures object functions, spatial relationships, and scene semantics, achieving cross-domain transfer through complete environment-self disentanglement; ii) a skill learning module for task execution, which processes self-state information including joint degrees of freedom and motor patterns, enabling cross-skill transfer through independent motor pattern encoding. We conducted extensive experiments on various LH tasks in HSI scenes. Compared with existing methods, ALAS can achieve an average subtasks success rate improvement of 23% and average execution efficiency improvement of 29%.

As terminal agents scale to long-horizon, multi-turn workflows, a key bottleneck is not merely limited context length, but the accumulation of noisy terminal observations in the interaction history. Retaining raw observations preserves useful environment feedback, but also leads to context saturation and high token cost; conversely, naive compression may discard task-critical signals needed for subsequent actions. Because terminal environments are highly heterogeneous across repositories, commands, and execution states, heuristic-based or fixed-prompt compression methods are difficult to generalize. We propose TACO, a plug-and-play, training-free, self-evolving Terminal Agent Compression framework for existing terminal agents. TACO automatically discovers, refines, and reuses structured compression rules from interaction trajectories, enabling workflow-adaptive filtering of low-value terminal outputs while preserving task-relevant observations. Experiments on TerminalBench (TB 1.0 and TB 2.0) and four additional terminal-related benchmarks, including SWE-Bench Lite, CompileBench, DevEval, and CRUST-Bench, show that TACO consistently improves task performance and token efficiency across agent scaffolds and backbone models. On TerminalBench, TACO yields 1%-4% accuracy gains across strong agentic models and improves accuracy by around 2%-3% under the same token budget. On additional terminal-related benchmarks, it reduces total token consumption while maintaining or improving task success rates. These results suggest that self-evolving, workflow-adaptive observation compression is an effective path toward more reliable and efficient long-horizon terminal agents. The code is publicly available at https://github.com/multimodal-art-projection/TACO.

In long-horizon open-world multi-agent systems, existing methods often treat local anomalies as automatic triggers for communication. This default design introduces coordination noise, interrupts local execution, and overuses public interaction in cases that could be resolved locally. To address this issue, we propose a partitioned information architecture for MLLM agents that explicitly separates private execution states from public coordination states. Building on this design, we introduce two key mechanisms. First, we develop an event-triggered working memory based on system-verified outcomes to maintain compact and low-noise local state representations. Second, we propose a cost-sensitive gated escalation mechanism that determines whether cross-region communication should be initiated by jointly considering node criticality, local recovery cost, and downstream task impact. In this way, communication is transformed from a default reaction into a selective decision. Experiments conducted on long-term construction tasks in open environments demonstrate that, compared to baseline models based on strong communication and planned structures, the introduction of gated communication and a partitioned information architecture results in superior performance in terms of blueprint completion quality and execution chain length. It also improves local self-recovery, reduces ineffective escalations, and increases the utility of public communication.

Long-horizon tool-using tasks sometimes benefit from revisiting earlier subtasks for recovery and exploration, but added multi-agent workflow flexibility can also introduce coordination overhead and substantial inference cost. We study complete cyclic subtask graphs, a deliberately maximally flexible multi-agent architecture in which executable subtask nodes are fully connected and a unified state-analysis-and-routing agent selects transitions using natural-language criteria. This makes unrestricted revisitation explicit and directly analyzable at the subtask level. We evaluate task-specific (Spec-Cyc) and benchmark-generic (Gen-Cyc) graphs on TextCraft, ALFWorld, and Finance-Agent, with ablations over planner/executor/router strength, tool exposure (generalist vs specialized), $n$-shot successful trajectory summaries, and fault-injected random subtask perturbations. The benchmarks expose three distinct regimes. ALFWorld highlights a setting where explicit revisitation supports recovery and exploration; TextCraft, a largely prerequisite-chain domain, often favors the efficiency of simpler forward execution; and Finance-Agent remains bottlenecked by retrieval, grounding, and evidence synthesis more than by workflow flexibility alone. Shared-win token comparisons further show that the added flexibility can be substantially more expensive than a single ReAct agent. Overall, we use complete cyclic subtask graphs as a maximally flexible experimental lens for measuring when multi-agent revisitation helps, when it mainly adds coordination cost, and when external task bottlenecks dominate.

LLM-based agent systems can adapt by revising prompts, constructing tools, and changing other runtime resources. Managing these heterogeneous updates requires a shared lifecycle for identifying targets, assessing candidates, and retaining accepted versions, complementing communication protocols such as A2A and MCP. We introduce AUTOGENESIS PROTOCOL (AGP), which decouples what evolves from how updates are proposed. Its Resource Substrate Protocol Layer (RSPL) represents prompts, agents, tools, environments, and memory as protocol-registered resources with explicit state and versioned interfaces. Its Self-Evolution Protocol Layer (SEPL) specifies candidate proposal, assessment, and commit, with auditable histories and support for restoring retained resource states. We instantiate AGP in AUTOGENESIS SYSTEM (AGS), a self-evolving multi-agent system that retrieves, instantiates, and refines resources during execution. Across reasoning, tool-use, and coding benchmarks, AGS improves over its non-evolving configuration, and repeated reasoning runs report the accompanying output-token cost. Execution cases show a shared update lifecycle across all five resource types.

Although artificial intelligence (AI) agents are increasingly proposed to support potentially longitudinal health tasks, such as symptom management, behavior change, and patient support, most current implementations fall short of facilitating user intent and fostering accountability. This contrasts with prior work on supporting longitudinal needs, both within and beyond clinical settings, where follow-up, coherent reasoning, and sustained alignment with individuals' goals are critical for both effectiveness and safety. In this paper, we draw on established clinical and personal health informatics frameworks to define what it would mean to orchestrate longitudinal health interactions with AI agents. We propose a multi-layer framework and corresponding agent architecture that operationalizes adaptation, coherence, continuity, and agency across repeated interactions. Through representative use cases, we demonstrate how longitudinal agents can maintain meaningful engagement, adapt to evolving goals, and support safe, personalized decision-making over time. Our findings underscore both the promise and the complexity of designing systems capable of supporting health trajectories beyond isolated interactions, and we offer guidance for future research and development in multi-session, user-centered health AI.

Large language model (LLM) agents perform strongly on short- and mid-horizon tasks, but often break down on long-horizon tasks that require extended, interdependent action sequences. Despite rapid progress in agentic systems, these long-horizon failures remain poorly characterized, hindering principled diagnosis and comparison across domains. To address this gap, we introduce HORIZON, an initial cross-domain diagnostic benchmark for systematically constructing tasks and analyzing long-horizon failure behaviors in LLM-based agents. Using HORIZON, we evaluate state-of-the-art (SOTA) agents from multiple model families (GPT-5 variants and Claude models), collecting 3100+ trajectories across four representative agentic domains to study horizon-dependent degradation patterns. We further propose a trajectory-grounded LLM-as-a-Judge pipeline for scalable and reproducible failure attribution, and validate it with human annotation on trajectories, achieving strong agreement (inter-annotator κ=0.61; human-judge κ=0.84). Our findings offer an initial methodological step toward systematic, cross-domain analysis of long-horizon agent failures and offer practical guidance for building more reliable long-horizon agents. We release our project website at HORIZON Leaderboard and welcome contributions from the community.

We study parallel test-time scaling for long-horizon agentic tasks such as agentic search and deep research, where multiple rollouts are generated in parallel and aggregated into a final response. While such scaling has proven effective for chain-of-thought reasoning, agentic tasks pose unique challenges: trajectories are long, multi-turn, and tool-augmented, and outputs are often open-ended. Aggregating only final answers discards rich information from trajectories, while concatenating all trajectories exceeds the model's context window. To address this, we propose AggAgent, an aggregation agent that treats parallel trajectories as an environment. We equip it with lightweight tools to inspect candidate solutions and search across trajectories, enabling it to navigate and synthesize information on demand. Across six benchmarks and three model families (GLM-4.7, Qwen3.5, MiniMax-M2.5), AggAgent outperforms all existing aggregation methods-by up to 5.3% absolute on average and 10.3% on two deep research tasks-while adding minimal overhead, as the aggregation cost remains bounded by a single agentic rollout. Our findings establish agentic aggregation as an effective and cost-efficient approach to parallel test-time scaling.

Enabling large language models to scale and reliably use hundreds of tools is critical for real-world applications, yet challenging due to the inefficiency and error accumulation inherent in flat tool-calling architectures. To address this, we propose Hybrid Toolset Agentization & Adaptation (HTAA), a hierarchical framework for scalable tool-use planning. We propose a novel toolset agentization paradigm, which encapsulates frequently co-used tools into specialized agent tools, thereby reducing the planner's action space and mitigating redundancy. To ensure effective coordination, we design Asymmetric Planner Adaptation, a trajectory-based training paradigm that aligns the high-level planner with agent tools via backward reconstruction and forward refinement. To validate the performance of HTAA, we conduct experiments on a real-world internal dataset, InfoVerify, based on the POI validation workflow of China's largest online large-scale ride-hailing platform, featuring long-horizon executable tool trajectories. Experiments on InfoVerify and widely-used benchmarks show that HTAA consistently achieves higher task success rates, requires short tool calling trajectories, and significantly reduces context overhead compared to strong baselines. Furthermore, in a production deployment, HTAA substantially reduces manual validation effort and operational cost, demonstrating its practical efficacy.

As large language models (LLMs) evolve into autonomous agents for long-horizon information-seeking, managing finite context capacity has become a critical bottleneck. Existing context management methods typically commit to a single fixed strategy throughout the entire trajectory. Such static designs may work well in some states, but they cannot adapt as the usefulness and reliability of the accumulated context evolve during long-horizon search. To formalize this challenge, we introduce a probabilistic framework that characterizes long-horizon success through two complementary dimensions: search efficiency and terminal precision. Building on this perspective, we propose AgentSwing, a state-aware adaptive parallel context management routing framework. At each trigger point, AgentSwing expands multiple context-managed branches in parallel and uses lookahead routing to select the most promising continuation. Experiments across diverse benchmarks and agent backbones show that AgentSwing consistently outperforms strong static context management methods, often matching or exceeding their performance with up to $3\times$ fewer interaction turns while also improving the ultimate performance ceiling of long-horizon web agents. Beyond the empirical gains, the proposed probabilistic framework provides a principled lens for analyzing and designing future context management strategies for long-horizon agents.

Multimodal agentic pipelines are transforming human-computer interaction by enabling efficient and accessible automation of complex, real-world tasks. However, recent efforts have focused on short-horizon or general-purpose applications (e.g., mobile or desktop interfaces), leaving long-horizon automation for domain-specific systems, particularly in healthcare, largely unexplored. To address this, we introduce CareFlow, a high-quality human-annotated benchmark comprising complex, long-horizon software workflows across medical annotation tools, DICOM viewers, EHR systems, and laboratory information systems. On this benchmark, existing vision-language models (VLMs) perform poorly, struggling with long-horizon reasoning and multi-step interactions in medical contexts. To overcome this, we propose CarePilot, a multi-agent framework based on the actor-critic paradigm. The Actor integrates tool grounding with dual-memory mechanisms (long-term and short-term experience) to predict the next semantic action from the visual interface and system state. The Critic evaluates each action, updates memory based on observed effects, and either executes or provides corrective feedback to refine the workflow. Through iterative agentic simulation, the Actor learns to perform more robust and reasoning-aware predictions during inference. Our experiments show that CarePilot achieves state-of-the-art performance, outperforming strong closed-source and open-source multimodal baselines by approximately 15.26% and 3.38%, respectively, on our benchmark and out-of-distribution dataset.

AI agents have become increasingly capable at isolated software engineering (SWE) tasks such as resolving issues on Github. Yet long-horizon tasks involving multiple interdependent subtasks still pose challenges both with respect to accuracy, and with respect to timely completion. A natural approach to solving these long-horizon tasks in a timely manner is asynchronous multi-agent collaboration, where multiple agents work on different parts of the task at the same time. But effective application of multi-agent systems has proven surprisingly difficult: concurrent edits by multiple agents interfere with each other, dependencies are difficult to synchronize, and combining partial progress into a coherent whole is challenging. On the other hand, human developers have long relied on mature collaboration infrastructure to manage these challenges in large software projects. Inspired by these collaboration primitives, we introduce Centralized Asynchronous Isolated Delegation (CAID), a structured multi-agent coordination paradigm grounded in three core SWE primitives: centralized task delegation, asynchronous execution, and isolated workspaces. CAID constructs dependency-aware task plans through a central manager, executes subtasks concurrently in isolated workspaces, and consolidates progress via structured integration with executable test-based verification. In empirical evaluation, we find that CAID improves accuracy over single-agent baselines by 25.6% absolute on paper reproduction tasks (PaperBench) and 14.7% on Python library development tasks (Commit0). Through systematic analysis, we find that branch-and-merge is a central coordination mechanism for multi-agent collaboration, and that SWE primitives such as git worktree, git commit, and git merge enable it to be realized in a reliable and executable manner.

Long-horizon embodied intelligence requires agents to improve through interaction, not merely to execute plans generated from static goals. A central challenge is therefore to transform past executions into knowledge that can shape future decisions. Minecraft provides a representative testbed for this problem, where tasks such as crafting tools, building redstone components, and obtaining diamond equipment involve long prerequisite chains and are frequently disrupted by missing tools, blocked paths, GUI failures, or stagnant execution. To this end, we propose MineEvolve, a knowledge-driven self-evolution framework that converts execution feedback into actionable behavioral knowledge. MineEvolve first uses \underline{\emph{\ding{182}Monitor}} to convert each subgoal execution into typed feedback, including state changes, inventory changes, failure types, progress signals, and stagnation indicators. \underline{\emph{\ding{183}Inducer}} then derives reusable skills from successful executions and remedies from failed or stagnant executions. \underline{\emph{\ding{184}Curator}} validates, merges, filters, and retrieves these knowledge entries, while \underline{\emph{\ding{185}Adaptor}} uses them to repair the unfinished part of the plan under repeated failures or stagnation. Experiments on the Minecraft MCU long-horizon task suite show that MineEvolve consistently improves performance across multiple language-model planners, with larger gains on high-dependency task groups. Ablation and knowledge-accumulation studies further demonstrate that converting execution signals into structured behavioral knowledge is an effective path toward self-evolving embodied agents in long-horizon environments. Our code is available at https://github.com/xzw-ustc/MC-MineEvolve.