# ROVER: Recursive Reasoning Over Videos with Vision-Language Models for Embodied Tasks

arXiv: https://arxiv.org/abs/2508.01943 · HTML: https://mlsanity.com/paper/2508.01943/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning.
- Builds on 37 of its 66 references that are in the map.
- First posted on arXiv 3 Aug 2025.

## Placed in

- LLM reasoning / [Spatial and embodied reasoning](https://mlsanity.com/lr/spatial-and-embodied-reasoning/)
- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 30 of 37)

- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control](https://mlsanity.com/paper/2307.15818/) (2023)
- [Do As I Can, Not As I Say: Grounding Language in Robotic Affordances](https://mlsanity.com/paper/2204.01691/) (2022)
- [RT-1: Robotics Transformer for Real-World Control at Scale](https://mlsanity.com/paper/2212.06817/) (2022)
- [Eureka: Human-Level Reward Design via Coding Large Language Models](https://mlsanity.com/paper/2310.12931/) (2023)
- [robosuite: A Modular Simulation Framework and Benchmark for Robot Learning](https://mlsanity.com/paper/2009.12293/) (2020)
- [RoboCasa: Large-Scale Simulation of Everyday Tasks for Generalist Robots](https://mlsanity.com/paper/2406.02523/) (2024)
- [VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Training](https://mlsanity.com/paper/2210.00030/) (2022)
- [MimicGen: A Data Generation System for Scalable Robot Learning using Human Demonstrations](https://mlsanity.com/paper/2310.17596/) (2023)
- [Language to Rewards for Robotic Skill Synthesis](https://mlsanity.com/paper/2306.08647/) (2023)
- [RoboNet: Large-Scale Multi-Robot Learning](https://mlsanity.com/paper/1910.11215/) (2019)
- [Bridge Data: Boosting Generalization of Robotic Skills with Cross-Domain Datasets](https://mlsanity.com/paper/2109.13396/) (2021)
- [ADaPT: As-Needed Decomposition and Planning with Language Models](https://mlsanity.com/paper/2311.05772/) (2023)
- [Text2Reward: Reward Shaping with Language Models for Reinforcement Learning](https://mlsanity.com/paper/2309.11489/) (2023)
- [RL-VLM-F: Reinforcement Learning from Vision Language Foundation Model Feedback](https://mlsanity.com/paper/2402.03681/) (2024)
- [Task and Motion Planning with Large Language Models for Object Rearrangement](https://mlsanity.com/paper/2303.06247/) (2023)
- [Reinforcement and Imitation Learning for Diverse Visuomotor Skills](https://mlsanity.com/paper/1802.09564/) (2018)
- [Learning Generalizable Robotic Reward Functions from "In-The-Wild" Human Videos](https://mlsanity.com/paper/2103.16817/) (2021)
- [Instruct2Act: Mapping Multi-modality Instructions to Robotic Actions with Large Language Model](https://mlsanity.com/paper/2305.11176/) (2023)
- [C-Learning: Learning to Achieve Goals via Recursive Classification](https://mlsanity.com/paper/2011.08909/) (2020)
- [Motif: Intrinsic Motivation from Artificial Intelligence Feedback](https://mlsanity.com/paper/2310.00166/) (2023)
- [On Bringing Robots Home](https://mlsanity.com/paper/2311.16098/) (2023)
- [Large Language Models as Generalizable Policies for Embodied Tasks](https://mlsanity.com/paper/2310.17722/) (2023)
- [Vision-Language Models as a Source of Rewards](https://mlsanity.com/paper/2312.09187/) (2023)
- [Unsupervised Perceptual Rewards for Imitation Learning](https://mlsanity.com/paper/1612.06699/) (2016)
- [Language Models as Zero-Shot Trajectory Generators](https://mlsanity.com/paper/2310.11604/) (2023)
- [TDAG: A Multi-Agent Framework based on Dynamic Task Decomposition and Agent Generation](https://mlsanity.com/paper/2402.10178/) (2024)

## Built on by (most cited first, 6 of 6)

- [Recurrent Reasoning with Vision-Language Models for Estimating Long-Horizon Embodied Task Progress](https://mlsanity.com/paper/2603.17312/) (2026)
- [SAIL: Test-Time Scaling for In-Context Imitation Learning with VLM](https://mlsanity.com/paper/2603.08269/) (2026)
- [SERUM: State Extraction and Refinement for User Modeling](https://mlsanity.com/paper/2607.29181/) (2026)
- [StageGuard: Learning Stage Transitions for Long-Horizon Robot Tasks via Agentic Distillation](https://mlsanity.com/paper/2609.20791/) (2026)
- [When Video Misreads: Closed-Loop Distillation of Reading Heuristics for Exploratory Manipulation Trace QA](https://mlsanity.com/paper/2606.08542/) (2026)
- [ARS: Agentic Reward System for Robot Learning](https://mlsanity.com/paper/2609.34484/) (2026)

## Abstract (arXiv)

Vision-language models (VLMs) have exhibited impressive capabilities across diverse image understanding tasks, but still struggle in settings that require reasoning over extended sequences of camera frames from a video. This limits their utility in embodied settings, which require reasoning over long frame sequences from a continuous stream of visual input at each moment of a task attempt. To address this limitation, we propose ROVER (Reasoning Over VidEo Recursively), a framework that enables the model to recursively decompose long-horizon video trajectories into segments corresponding to shorter subtasks within the trajectory. In doing so, ROVER facilitates more focused and accurate reasoning over temporally localized frame sequences without losing global context. We evaluate ROVER, implemented using an in-context learning approach, on diverse OpenX Embodiment videos and on a new dataset derived from RoboCasa that consists of 543 videos showing both expert and perturbed non-expert trajectories across 27 robotic manipulation tasks. ROVER outperforms strong baselines across three video reasoning tasks: task progress estimation, frame-level natural language reasoning, and video question answering. We observe that, by reducing the number of frames the model reasons over at each timestep, ROVER mitigates hallucinations, especially during unexpected or non-optimal moments of a trajectory. In addition, by enabling the implementation of a subtask-specific sliding context window, ROVER's time complexity scales linearly with video length, an asymptotic improvement over baselines. Demos, code, and data available at: https://rover-vlm.github.io