Research map / Vision-language and multimodal / Robot action and navigation
mVLM: A Vision Language Model for mNPUs
Zijie Chen, Guiyun Fan, Zhaoxing Yang, Rong Ding, Haiming Jin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add dynamic ED blocks to OverMod encoder (ablation) | OverMod · ImageNet-1K val | top-1 accuracy (%): 75.8 -> 77.6 | Improved |
| add Dilated RepConv to OverMod (ablation) | OverMod · ImageNet-1K val | top-1 accuracy (%): 77.6 -> 78.4 | Improved |
| add GRN to OverMod (ablation) | OverMod · ImageNet-1K val | top-1 accuracy (%): 78.4 -> 78.9 | Improved |
| add Layer Scale to OverMod (ablation) | OverMod · ImageNet-1K val | top-1 accuracy (%): 78.9 -> 79.2 | Improved |
| add SE module to OverNet or ModNet encoder (ablation) | OverMod · ImageNet-1K val | top-1 accuracy (%): 79.2 -> 79.0 | Hurt |
| add dynamic convolution (ablation) | µVLM · COCO Karpathy test | CIDEr: 95.5 -> 101.4 | Improved |
| add multi-scale feature fusion (ablation) | µVLM · COCO Karpathy test | CIDEr: 101.4 -> 107.1 | Improved |
| add spatial modulation in decoder attention (ablation) | µVLM · COCO Karpathy test | CIDEr: 107.1 -> 115.2 | Improved |
| add channel modulation in decoder attention (ablation) | µVLM · COCO Karpathy test | CIDEr: 115.2 -> 117.8 | Improved |
| replace LSTM with Bahdanau attention by AttSSM decoder (ablation) | µVLM on STM32N657 · on-device latency | latency (ms): 32 -> 21 | Improved |
| scale OverMod from tiny to base | µVLM · COCO Karpathy test | CIDEr: 96.4 -> 117.8 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- SmallCap — SmallCap: Lightweight Image Captioning Prompted With Retrieval Augmentation (CVPR 2023) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- VLsI: Verbalized Layers-to-Interactions from Large to Small Vision Language Models (CVPR 2025 · similarity 0.87)
- Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Models (ICML 2024 · similarity 0.86)
- Dynamic-LLaVA: Efficient Multimodal Large Language Models via Dynamic Vision-language Context Sparsification (ICLR 2025 · similarity 0.86)
- What matters when building vision-language models? (NeurIPS 2024 · similarity 0.85)
- ViSpec: Accelerating Vision-Language Models with Vision-Aware Speculative Decoding (NeurIPS 2025 · similarity 0.85)
- MHA2MLA-VLM: Enabling DeepSeek’s Economical Multi-Head Latent Attention Across Vision-Language Models (AAAI 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
