# RetryRoute: Retry-Aware Model Routing for Agentic Image Editing Services

DevFeed: [RetryRoute: Retry-Aware Model Routing for Agentic Image Editing Services](<https://devfeed.tech/articles/retryroute-retry-aware-model-routing-for-agentic-image-editing-services-111589.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.41460>)

Author: Hanshuai Cui, Zhiqing Tang, Zhi Yao, Jun Luo, Weijia Jia

Published: 2026-11-02T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Model Routing](<https://devfeed.tech/topics/model-routing.md>), [retry](<https://devfeed.tech/topics/retry.md>), [image editing](<https://devfeed.tech/topics/image-editing.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [dispatch](<https://devfeed.tech/tags/dispatch.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [image](<https://devfeed.tech/tags/image.md>), [image-editing](<https://devfeed.tech/tags/image-editing.md>), [model-routing](<https://devfeed.tech/tags/model-routing.md>), [retry](<https://devfeed.tech/tags/retry.md>), [routing](<https://devfeed.tech/tags/routing.md>)

## AI overview

Retries after a rejected image edit are less likely to succeed than first attempts: across six open models, three benchmarks and two judges, retry success was only 17-53% of first-try success. RetryRoute uses each request's failure history to estimate model success and plan retries, escalations or abandonment, incorporating live queue waits into its cost. At the main budget, it beat the strongest of eleven prior baselines by an average 2.8 points in success under the in-loop judge across eight groups. On a GPU testbed, it increased edits delivered within the service-level objective by 5.0-6.3 points over the best prior baseline at every tested load.

## Source excerpt

arXiv:2610.41460v1 Announce Type: new Abstract: Web agents increasingly edit images by calling open editing models in a critique-and-retry loop, and these models differ in cost by more than an order of magnitude. After every rejected edit, a service must decide whether to retry the same model, escalate to another or give up. Existing routers pick one model per request, and cascades escalate in a fixed order. However, both treat a retry as an independent draw. We identify sticky failures, meaning that a retry after a failure succeeds only 17-53% as often as a first try across six open models, three benchmarks and two judges. Based on this finding, we propose RetryRoute, a retry-aware router that estimates each model's chance of success given the request's failure history. It plans retries, escalations and abandonment over failure states and adds live queue waits to the cost, so the same plan becomes load-aware dispatch. At the main budget, RetryRoute's success under the in-loop judge exceeds the strongest of eleven prior baselines by 2.8 points on average over eight groups. On a GPU serving testbed it raises the share of edits delivered within the service-level objective by 5.0-6.3 points over the best prior baseline at every load.