COMAP: Co-Evolving World Models and Agent Policies for LLM Agents

• URL: http://arxiv.org/abs/2606.02372v1
• Date: Mon, 01 Jun 2026 15:21:17 GMT
• Title: COMAP: Co-Evolving World Models and Agent Policies for LLM Agents
• Abstract summary: COMAP is a novel framework that co-evolves textual world models and agent policies through closed-loop interaction.<n>At each decision step, the world model predicts future state feedback for candidate actions, and the agent performs future-aware reflection by estimating the reliability of this feedback.<n>The resulting on-policy trajectories are then used to update the world model via self-distillation, allowing it to better match the agent's evolving interaction distribution.
• Score: 14.918267305899619
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Equipping language agents with world models enables them to anticipate environment dynamics and evaluate candidate actions before execution. However, existing textual world models are typically fixed after training, preventing them from adapting to the on-policy state-action distributions induced by an evolving agent. Meanwhile, agent-improvement methods often rely on external rewards or verifiers, limiting their applicability in realistic interactive environments. In this paper, we propose COMAP, a novel framework that co-evolves textual world models and agent policies through closed-loop interaction. At each decision step, the world model predicts future state feedback for candidate actions, and the agent performs future-aware reflection by estimating the reliability of this feedback and refining its action accordingly. The resulting on-policy trajectories are then used to update the world model via self-distillation, allowing it to better match the agent's evolving interaction distribution. Across embodied task planning, Web navigation, and tool-use benchmarks, COMAP consistently outperforms competitive baselines, e.g., +16.75% relative improvement with Qwen3-4B. Further analyses show that the co-evolutionary loop improves the world model's prediction accuracy over time and leads to more effective long-horizon decision-making. Our code is available at: https://github.com/loyiv/CoMAP.

Related papers

• Quo Vadis, World Modeling? [116.2429173084643]
  Continually improving agents require dynamic interaction feedback beyond static supervision, yet direct real-environment interaction is costly, slow, unsafe, and hard to parallelize.<n>World modeling offers a natural intermediate proxy that allows agents to query lower-cost, more controllable feedback before committing to real actions.<n>This work recasts world modeling into an agent-centric paradigm, establishing a roadmap for building world proxies that empower agents to plan better, learn faster, and evolve continually.
  arXiv  Detail & Related papers  (2026-08-03T17:59:58Z)
• From World Action Models to Embodied Brains: A Roadmap for Open-World Physical Intelligence [54.62791372989525]
  Action models, vision-language--action policies, and world models have advanced this goal.<n>World Action Models (WAMs) are particularly promising because they connect candidate interventions with predicted consequences.<n>We review the evolution toward WAMs and organize these limitations into three coupled gaps: model roles and representations, objectives and standardization, and system composition.
  arXiv  Detail & Related papers  (2026-07-13T15:22:56Z)
• Self-Evolving World Models for LLM Agent Planning [72.99782619992361]
  We introduce WorldEvolver, a self-evolving world model framework that revises its deployment-time context while keeping the downstream agent and all model parameters frozen.<n>WorldEvolver achieves the highest prediction accuracy across three backbones and leads other world model baselines on downstream agent success rate.
  arXiv  Detail & Related papers  (2026-06-29T17:58:43Z)
• Autoregressive Diffusion World Models for Off-Policy Evaluation of LLM Agents [35.37126280179998]
  evaluating large language model (LLM) agents in multi-turn interactive environments is expensive and risky.<n>We propose ADWM, an evaluation framework that estimates the performance of a new LLM agent policy purely from pre-collected trajectories.
  arXiv  Detail & Related papers  (2026-06-04T01:13:24Z)
• How Mobile World Model Guides GUI Agents? [51.29718003718467]
  We train world models across four modalities: delta text, full text, diffusion-based images, and renderable code.<n>These models achieve SoTA performance on both MobileWorldBench and Code2WorldBench.<n>For overconfident mobile agents with low action entropy, posterior self-reflection provides limited gains.
  arXiv  Detail & Related papers  (2026-05-11T10:49:31Z)
• Beyond State Consistency: Behavior Consistency in Text-Based World Models [58.216587360435305]
  We introduce a new behavior-aligned training paradigm aimed at improving the functional consistency between the world model and the real environment.<n> Experiments on WebShop and TextWorld show that BehR-based training improves long-term alignment in several settings.
  arXiv  Detail & Related papers  (2026-04-15T12:56:45Z)
• World Action Verifier: Self-Improving World Models via Forward-Inverse Asymmetry [82.93104394404781]
  General-purpose world models promise scalable policy evaluation, optimization, and planning.<n>We propose World Action Verifier (WAV), a framework that enables world models to identify their own prediction errors and self-improve.
  arXiv  Detail & Related papers  (2026-04-02T12:48:36Z)
• Imagine-then-Plan: Agent Learning from Adaptive Lookahead with World Models [13.90532093512575]
  We propose Imagine-then-Plan (textttITP), a unified framework for agent learning via lookahead imagination.<n>We introduce a novel adaptive lookahead mechanism by trading off the ultimate goal and task progress.<n>Our experiments show that textttITP significantly outperforms competitive baselines.
  arXiv  Detail & Related papers  (2026-01-13T19:49:58Z)
• AlignUSER: Human-Aligned LLM Agents via World Models for Recommender System Evaluation [0.7031557790463293]
  We introduce Align, a framework that learns world-model-driven agents from human interactions.<n>We generate counterfactual trajectories around demonstrations and prompt the LLM to compare its decisions with human choices, identify suboptimal actions, and extract lessons.
  arXiv  Detail & Related papers  (2026-01-02T03:01:33Z)
• Agent2World: Learning to Generate Symbolic World Models via Adaptive Multi-Agent Feedback [51.22403664895878]
  Agent2World is a tool-augmented multi-agent framework that achieves strong inference-time world-model generation.<n>It also serves as a data engine for supervised fine-tuning, by grounding generation in multi-agent feedback.
  arXiv  Detail & Related papers  (2025-12-26T18:54:14Z)
• WorldGym: World Model as An Environment for Policy Evaluation [41.204900701616914]
  WorldGym is an autoregressive, action-conditioned video generation model which serves as a proxy to real world environments.<n> Policies are evaluated via Monte Carlo rollouts in the world model, with a vision-language model providing rewards.<n>We show that WorldGym is able to preserve relative policy rankings across different policy versions, sizes, and training checkpoints.
  arXiv  Detail & Related papers  (2025-05-31T15:51:56Z)
• World Models via Policy-Guided Trajectory Diffusion [21.89154719069519]
  Existing world models are autoregressive in that they interleave predicting the next state with sampling the next action from the policy. We propose a novel world modelling approach that is not autoregressive and generates entire on-policy trajectories in a single pass through a diffusion model.
  arXiv  Detail & Related papers  (2023-12-13T21:46:09Z)
• Model-based Reinforcement Learning for Decentralized Multiagent Rendezvous [66.6895109554163]
  Underlying the human ability to align goals with other agents is their ability to predict the intentions of others and actively update their own plans. We propose hierarchical predictive planning (HPP), a model-based reinforcement learning method for decentralized multiagent rendezvous.
  arXiv  Detail & Related papers  (2020-03-15T19:49:20Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.