# CoCoSI: Collaborative Cognitive Map Construction for Spatial Intelligence

Source: [CoCoSI: Collaborative Cognitive Map Construction for Spatial Intelligence](<https://www.sota2.com/research/paper/6a28f32090966ade919c1637>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-12T05:00:05.419Z

Authors: Yiming Zhang, Ruoxuan Cao, Zhihang Zhong

Published: 2026-06-09T04:20:08.000Z

## Abstract

Spatial intelligence is a key frontier for multimodal large language models (MLLMs), enabling them to reason about the physical world from visual experience. Inspired by human spatial cognition, recent approaches construct grid-based cognitive maps from multi-frame visual inputs to maintain coherent spatial representations over time. However, limited context lengths still challenge spatial understanding, while existing methods, such as long-context modeling and external memory, often require architectural changes, memory modules, or finetuning, limiting their applicability to off-the-shelf pretrained MLLMs. This motivates a lightweight, model-agnostic method for preserving spatial information beyond the native context window. To this end, we propose a plug-and-play multi-agent framework that collaboratively constructs cognitive maps as structured spatial memory, enhancing the spatial understanding of arbitrary pretrained MLLMs without architectural modification or additional training. Our framework features local-global agent coordination, cognitive map construction with atomic commits, and cross-agent verification. Extensive experiments demonstrate that our method achieves superior performance on spatial understanding tasks while remaining fully training-free. Code will be released.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Video Spatial Understanding on VSI-Bench 500 QA 1.0](<https://www.sota2.com/research/sota/video-spatial-understanding-on-vsi-bench-500-qa-1-0.md>): Video Spatial Understanding; VSI-Bench 500 QA 1.0; Accuracy: 49.7.

- [Video Spatial Understanding on MMSI-Bench 500 QA 1.0](<https://www.sota2.com/research/sota/video-spatial-understanding-on-mmsi-bench-500-qa-1-0.md>): Video Spatial Understanding; MMSI-Bench 500 QA 1.0; Accuracy: 41.

- [Video Spatial Understanding on MMSI-Bench 500 QA subset](<https://www.sota2.com/research/sota/video-spatial-understanding-on-mmsi-bench-500-qa-subset.md>): Video Spatial Understanding; MMSI-Bench 500 QA subset; Accuracy: 35.3.

- [Video Spatial Understanding on MMSI-Bench 500 QA pairs](<https://www.sota2.com/research/sota/video-spatial-understanding-on-mmsi-bench-500-qa-pairs.md>): Video Spatial Understanding; MMSI-Bench 500 QA pairs; Accuracy: 41.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2606.10401>)

[PDF](<https://arxiv.org/pdf/2606.10401.pdf>)