RoboTALES: Learning Reasoning-Guided Robot Policies via Task-Aligned Simulated Futures

2026-07-07

AI Summary

Pretrained video generation models often struggle in robotic control due to future predictions that deviate from task intent and unreliable action conditioning, hindering effective policy learning. To address this, this work proposes RoboTALES, a framework that leverages a hierarchical large language model (LLM) to decompose tasks into subgoals, thereby guiding the video generation model to imagine task-aligned future trajectories. Additionally, a vision-language model (VLM) acts as a critic, utilizing reward feedback to ensure the generated trajectories remain consistent with the intended task objective, enabling reasoning-guided end-to-end policy learning. Evaluated on the RoboCasa and LIBERO10 benchmarks, RoboTALES significantly outperforms existing methods, demonstrating superior temporal consistency and task success rates, particularly in long-horizon manipulation tasks.

Abstract

Pretrained video generative models are promising backbones for visuomotor control, but their imagined futures often drift from task intent and are not reliably action-conditional. As a result, these models can be difficult to use for planning or policy extraction. To address these limitations, we propose RoboTALES, a single-stage framework that learns task-aligned simulated futures and uses them to train robot policies. Our approach introduces two key innovations: (1) a hierarchical LLM-based planner that breaks complex tasks into a sequence of subgoals to guide the model's imagination; and (2) a VLM-based critic that evaluates these ``imagined'' futures and uses reward-based feedback to keep the model's internal representations focused on the goal. By anchoring the video generator in abstract reasoning, we produce temporally consistent rollouts and more coherent actions. We evaluate RoboTALES on diverse manipulation tasks from RoboCasa and LIBERO10, and show that our method consistently outperforms existing methods, especially in long-horizon tasks. Our code and models are publicly available at https://github.com/hananshafi/RoboTALES.

Problem

• video generative models
• visuomotor control
• task-aligned futures
• action-conditionality
• policy extraction

Innovation

• reasoning-guided planning
• task-aligned video generation
• hierarchical LLM planner
• VLM-based critic
• visuomotor policy learning