mlsanityData to 14 Oct 2026

Distilling Vision-Language Models for On-Device Fire Understanding

arXiv 2609.05782 ↗First posted 5 Sep 2026cs.AIcited by 0

On the map: Multimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness

also in 1 more
Abstract (arXiv)

Vision-language models (VLMs) offer a promising alternative to conventional fire detection systems by reasoning about the semantic context of a scene and thus reducing false alarms, yet their large model size makes deployment on embedded fire sensors impractical. In this paper, we study how domain-specialized VLMs can be compressed for fully on-device deployment without losing the safety-critical behavior required for fire detection. We develop a teacher-student knowledge distillation framework in which large VLMs fine-tuned for fire understanding can be distilled into lightweight students. Experiments across multiple VLM families and model scales show that compact students preserve most of their teachers' fire-understanding capability. We further deploy the distilled models on our commercial Detectium fire detection sensor and jointly evaluate reasoning accuracy, latency, and memory usage. The results show that compression and deployment affect not only accuracy but also model failure modes, with Qwen2.5-0.5B providing the strongest overall deployment trade-off. Our findings provide broader guidance for deploying domain-specialized VLMs in resource-constrained, safety-critical settings.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 11 of its 31 references that are in the map. First posted on arXiv 5 Sep 2026.

Builds on

11 of 31 references.

References · 11

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  3. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  4. 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
  5. 2022Reproducible scaling laws for contrastive language-image learningcited by 1,173
  6. 2016Sequence-Level Knowledge Distillationcited by 801
  7. 2023FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterizationcited by 162
  8. 2023MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Trainingcited by 96
  9. 2024LLaVA-MoD: Making LLaVA Tiny via MoE Knowledge Distillationcited by 45
  10. 2025SmokeBench: Evaluating Multimodal Large Language Models for Wildfire Smoke Detectioncited by 5
  11. 2025DetectiumFire: A Comprehensive Multi-modal Dataset Bridging Vision and Language for Fire Understandingcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026SAFIRE: Safety-Critical Benchmark for Fine-grained Fire and Smoke Understanding in Multimodal LLMs
  2. 2026ThermEval: A Structured Benchmark for Evaluation of Vision-Language Models on Thermal Imagery
  3. 2025Rethinking Bottlenecks in Safety Fine-Tuning of Vision Language Models
  4. 2025iSafetyBench: A video-language benchmark for safety in industrial environment
  5. 2025Real-Time Aerial Fire Detection on Resource-Constrained Devices Using Knowledge Distillation
  6. 2025Robotic Fire Risk Detection based on Dynamic Knowledge Graph Reasoning: An LLM-Driven Approach with Graph Chain-of-Thought
  7. 2026Hazard or Anomaly? Evaluating VLMs for Understanding Dangers and Discrepancies
  8. 2026ThermoAct:Thermal-Aware Vision-Language-Action Models for Robotic Perception and Decision-Making