# CORA: Conformal Risk-Controlled Agents for Safeguarded Mobile GUI Automation

Source: [CORA: Conformal Risk-Controlled Agents for Safeguarded Mobile GUI Automation](<https://www.sota2.com/research/paper/69dc786b7d0172140ec789a2>)

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

Data updated: 2026-04-14T05:00:04.234Z

Authors: Yushi Feng, Junye Du, Qifan Wang, Zizhan Ma, Qian Niu, Yutaka Matsuo, Long Feng, Lequan Yu

Published: 2026-04-10T09:41:21.000Z

## Abstract

Graphical user interface (GUI) agents powered by vision language models (VLMs) are rapidly moving from passive assistance to autonomous operation. However, this unrestricted action space exposes users to severe and irreversible financial, privacy or social harm. Existing safeguards rely on prompt engineering, brittle heuristics and VLM-as-critic lack formal verification and user-tunable guarantees. We propose CORA (COnformal Risk-controlled GUI Agent), a post-policy, pre-action safeguarding framework that provides statistical guarantees on harmful executed actions. CORA reformulates safety as selective action execution: we train a Guardian model to estimate action-conditional risk for each proposed step. Rather than thresholding raw scores, we leverage Conformal Risk Control to calibrate an execute/abstain boundary that satisfies a user-specified risk budget and route rejected actions to a trainable Diagnostician model, which performs multimodal reasoning over rejected actions to recommend interventions (e.g., confirm, reflect, or abort) to minimize user burden. A Goal-Lock mechanism anchors assessment to a clarified, frozen user intent to resist visual injection attacks. To rigorously evaluate this paradigm, we introduce Phone-Harm, a new benchmark of mobile safety violations with step-level harm labels under real-world settings. Experiments on Phone-Harm and public benchmarks against diverse baselines validate that CORA improves the safety--helpfulness--interruption Pareto frontier, offering a practical, statistically grounded safety paradigm for autonomous GUI execution. Code and benchmark are available at cora-agent.github.io.

## 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.

- [Safe Navigation on AndroidWorld core20 safe general tasks](<https://www.sota2.com/research/sota/safe-navigation-on-androidworld-core20-safe-general-tasks.md>): Safe Navigation; AndroidWorld core20 safe general tasks; Success Count (out of 20): 8.

- [Mobile Agent Safety and Capability Evaluation on Phone-Harm + Normal-150 (merged)](<https://www.sota2.com/research/sota/mobile-agent-safety-and-capability-evaluation-on-phone-harm-plus-normal-150-merged.md>): Mobile Agent Safety and Capability Evaluation; Phone-Harm + Normal-150 (merged); Harm Rate (HR): 2.19.

- [Risk Detection on MobileRisk](<https://www.sota2.com/research/sota/risk-detection-on-mobilerisk.md>): Risk Detection; MobileRisk; Accuracy: 61.27.

- [Safety and Helpfulness Evaluation on Phone-Harm Harm-150](<https://www.sota2.com/research/sota/safety-and-helpfulness-evaluation-on-phone-harm-harm-150.md>): Safety & Helpfulness Evaluation; Phone-Harm Harm-150; HR: 4.37.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.09155>)

[PDF](<https://arxiv.org/pdf/2604.09155.pdf>)
