Saved in:
Bibliographic Details
Main Author: Zhou, Yunpeng
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.31354
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910272387547136
author Zhou, Yunpeng
author_facet Zhou, Yunpeng
contents Modular visual reasoning systems increasingly rely on shared working memory for multi-step collaboration, yet the failure dynamics of intermediate state evolution in low-capacity regimes remain underexplored. We study failure modes of collaborative reasoning with weak learners (4B--8B models) through the lens of noise accumulation. We introduce CoSee, an auditing framework that formalizes the read-write-verify loop to trace information flow in document visual question answering. Across multi-page, chart, and web-based benchmarks, we find a counter-intuitive degradation: naive shared workspaces often amplify hallucinations rather than resolve them. We identify two dominant failure modes: Noise Reinforcement, where ungrounded notes are reused as evidence, and Policy Collapse, where added context shifts the model toward under-specified, short-form answers. Using cost-accuracy Pareto frontiers, we show that increased compute can correlate negatively with performance without explicit verification. Our findings suggest that for resource-constrained agents, the bottleneck lies not in reasoning depth but in communication fidelity, providing trace-level diagnostics and a mechanistic baseline for reliable modular design.
format Preprint
id arxiv_https___arxiv_org_abs_2605_31354
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Diagnosing Failure Modes of Shared-State Collaboration in Resource-Constrained Visual Agents
Zhou, Yunpeng
Artificial Intelligence
Machine Learning
Modular visual reasoning systems increasingly rely on shared working memory for multi-step collaboration, yet the failure dynamics of intermediate state evolution in low-capacity regimes remain underexplored. We study failure modes of collaborative reasoning with weak learners (4B--8B models) through the lens of noise accumulation. We introduce CoSee, an auditing framework that formalizes the read-write-verify loop to trace information flow in document visual question answering. Across multi-page, chart, and web-based benchmarks, we find a counter-intuitive degradation: naive shared workspaces often amplify hallucinations rather than resolve them. We identify two dominant failure modes: Noise Reinforcement, where ungrounded notes are reused as evidence, and Policy Collapse, where added context shifts the model toward under-specified, short-form answers. Using cost-accuracy Pareto frontiers, we show that increased compute can correlate negatively with performance without explicit verification. Our findings suggest that for resource-constrained agents, the bottleneck lies not in reasoning depth but in communication fidelity, providing trace-level diagnostics and a mechanistic baseline for reliable modular design.
title Diagnosing Failure Modes of Shared-State Collaboration in Resource-Constrained Visual Agents
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2605.31354