GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
xue-26
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
xue-26
Model Tree
Input Modalities
Output Modalities
Supported Functionality
SeerGuard is designed as an external guard model, not as a replacement for the mobile GUI agent.
Clarification: SAWM itself is fine-tuned from Qwen3-VL-8B-Instruct. "No fine-tuning" means that deploying SeerGuard does not require additional training or weight updates for the GUI agent being protected.
Mobile GUI agents interact with live applications, accounts, data, and device settings. A single incorrect tap may trigger an unauthorized payment, expose private information, send harmful content, delete data, or change a security-sensitive configuration.
Conventional post-hoc safety checkers review an interaction after the action has already occurred. SeerGuard instead performs proactive consequence auditing:
text
GUI agent proposes an action|vSeerGuard predicts the semantic consequence|vSafety decision: safe or unsafe
This design enables the host agent to retain its original task-completion capabilities while SeerGuard independently handles safety assessment.
Before the GUI agent interacts with the device, SAWM determines whether the user request is explicitly malicious, unauthorized, or in conflict with core safety policies.
text
Input:- User instructionOutput:- safety-judge: safe or unsafe,- safety-reason: Brief explanation
Unsafe instructions are rejected before planning and execution.
For each candidate action, SAWM predicts the likely functional consequence in natural language. It focuses on the semantic state transition rather than generating the next screen at the pixel level.
Examples include:
SAWM evaluates whether the predicted consequence is safe and provides a rationale for the decision.
text
Input:- Current mobile GUI screenshot- User instruction- Candidate actionOutput:- Current screen analysis- Predicted semantic next state- Safety rationale- Action safety label: safe | unsafe
If the action is unsafe, SeerGuard blocks it before execution. Otherwise, the host agent proceeds normally.
The GUI agent remains responsible for perception, planning, and task completion. SeerGuard only decides whether a request or candidate action should be allowed to proceed.
SAWM uses Qwen3-VL-8B-Instruct as its backbone and is trained under a unified autoregressive formulation. It jointly learns:
For action-level assessment, the model returns a structured response:
json
{"ui_context_analysis": "...","action_consequence_prediction": "...","risk_evaluation": "...","safety_judgment": "SAFE" | "UNSAFE"}
SAWM is trained with a multi-task corpus that combines mobile world-modeling data with safety-augmentation data:
General Textual Safety Data
Multimodal Mobile Risk Data
MobileWorld Next-State QA Data
Synthetic Textual Mobile Risk Data
The final training corpus contains approximately 148K instances. SAWM is fine-tuned for 1 epoch with a learning rate of 1e-6.
Full evaluation code and benchmark instructions are available in the SeerGuard GitHub repository.
On MobileSafetyBench, SeerGuard improves the safety-utility trade-off across Qwen3-VL, GPT-5.1, and Gemini-3.1 GUI-agent backbones.
For Qwen3-VL at a safety-priority setting of 0.8:
SAWM is evaluated on Agent-SafetyBench and Prompt Injection benchmarks. On the Prompt Injection benchmark, SAWM achieves an F1 score of 0.922, demonstrating strong risk detection while limiting unnecessary refusal of safe requests.
On MobileRisk, SAWM achieves:
Both are the best results among the compared methods, showing that SAWM can identify unsafe trajectories and locate the onset of risk.
On Next-State-QA, SAWM reaches an accuracy of 0.762, providing the predictive grounding required for consequence-aware action auditing.
SAWM is intended for research and development of safety guardrails for mobile GUI agents.
Typical use cases include:
If you find this model useful, please cite the SeerGuard paper:
bibtex
@inproceedings{seerguard2026,title = {SeerGuard: A Safety Framework for Mobile GUI Agents via World Model Prediction},author = {Xue Yu and Bo Yuan and Pengshuai Yang and Kailin Zhao and Hong Hu and Junlan Feng},booktitle = {Proceedings of the ACM Multimedia Conference},year = {2026},note = {To appear}}
SAWM is built on Qwen3-VL-8B-Instruct. Please also cite the corresponding Qwen-VL technical report when using this model.
bibtex
@misc{qwen3technicalreport,title = {Qwen3 Technical Report},author = {Qwen Team},year = {2025},eprint = {2505.09388},archivePrefix = {arXiv},primaryClass = {cs.CL},url = {https://arxiv.org/abs/2505.09388}}