MonitorVLM-v2: A Deployed Vision-Language Framework for Real-Time Safety Violation Detection

2026-08-01

AI Summary

This work addresses the limitations of conventional vision-language models in industrial safety monitoring, where open-ended autoregressive inference fails to meet stringent requirements for real-time performance, multi-stream concurrency, and deterministic hazard detection. The authors reformulate safety assessment as probabilistic reasoning over a finite symbolic space, replacing variable-length chain-of-thought generation with single-step symbol prediction. They introduce Symbolic Policy Optimization (SymPO) to sharpen decision boundaries and integrate an entropy-driven uncertainty routing mechanism to enable efficient human-in-the-loop verification. Implemented with a single-token decoding architecture, the system achieves a 19.45× speedup in inference latency and detects 2.78× more confirmed violations than manual inspections in real-world underground mine deployments, effectively balancing efficiency, auditability, and detection performance.

Abstract

Large vision--language models (VLMs) can reason step by step about complex visual scenes, but this open-ended, autoregressive chain-of-thought (CoT) approach is poorly suited to safety-critical, rule-governed settings such as industrial surveillance, where decisions must be bounded, deterministic, and low-latency. Because CoT inference cost scales jointly with reasoning length and the number of concurrent streams, it creates a throughput bottleneck that precludes the real-time, multistream monitoring required for industrial accountability. Here we present MonitorVLM-v2, a deployment-oriented framework that recasts VLM-based safety assessment as probabilistic inference over a finite regulatory decision space, compressing multimodal reasoning into single-step rule-ID predictions and reducing decoding from a variable-length sequence to a single token. We introduce symbolic policy optimization (SymPO), a novel contrastive policy optimization algorithm that sharpens decision boundaries within this finite symbolic space, together with an entropy-driven triage mechanism that routes uncertain predictions to human reviewers for expert confirmation. In a four-month prospective deployment across 10 concurrent camera feeds in an operational underground mining facility, MonitorVLM-v2 achieved a 19.45-fold increase in inference speed and identified 2.78 times as many confirmed violations as the site's routine manual inspection workflow, demonstrating the practical value of compressed symbolic decision-making for real-time, auditable industrial monitoring.

Problem

• safety violation detection
• vision-language models
• real-time monitoring
• industrial surveillance
• low-latency inference

Innovation

• symbolic policy optimization
• finite decision space
• real-time multimodal monitoring
• entropy-driven triage
• single-token decoding