mlsanityData to 14 Oct 2026

Atom: Efficient On-Device Video-Language Pipelines Through Modular Reuse

arXiv 2512.17108 ↗First posted 18 Dec 2025cs.LG cs.MMcited by 0

Not on the map. Why?

Abstract (arXiv)

Recent advances in video-language models have enabled powerful applications like video retrieval, captioning, and assembly. However, executing such multi-stage pipelines efficiently on mobile devices remains challenging due to redundant model loads and fragmented execution. We introduce Atom, an on-device system that restructures video-language pipelines for fast and efficient execution. Atom decomposes a billion-parameter model into reusable modules, such as the visual encoder and language decoder, and reuses them across subtasks like captioning, reasoning, and indexing. This reuse-centric design eliminates repeated model loading and enables parallel execution, reducing end-to-end latency without sacrificing performance. On commodity smartphones, Atom achieves 27--33% faster execution compared to non-reuse baselines, with only marginal performance drop (≤ 2.3 Recall@1 in retrieval, ≤ 1.5 CIDEr in captioning). These results position Atom as a practical, scalable approach for efficient video-language understanding on edge devices.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 22 of its 37 references that are in the map. First posted on arXiv 18 Dec 2025.

Builds on

22 of 37 references.

References · 22

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  3. 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,298
  4. 2021Frozen in Time: A Joint Video and Image Encoder for End-to-End Retrievalcited by 1,177
  5. 2021CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retrievalcited by 651
  6. 2023NExT-GPT: Any-to-Any Multimodal LLMcited by 615
  7. 2024Panda-70M: Captioning 70M Videos with Multiple Cross-Modality Teacherscited by 393
  8. 2024InternVideo2: Scaling Foundation Models for Multimodal Video Understandingcited by 317
  9. 2024The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bitscited by 308
  10. 2023mPLUG-2: A Modularized Multi-modal Foundation Model Across Text, Image and Videocited by 154
  11. 2018Localizing Moments in Video with Temporal Languagecited by 83
  12. 2024LLM Pruning and Distillation in Practice: The Minitron Approachcited by 82
10 more
  1. 2022Spotlight: Mobile UI Understanding using Vision-Language Models with a Focuscited by 62
  2. 2024Fast On-device LLM Inference with NPUscited by 62
  3. 2025VideoRAG: Retrieval-Augmented Generation over Video Corpuscited by 57
  4. 2024Any-Precision LLM: Low-Cost Deployment of Multiple, Different-Sized LLMscited by 54
  5. 2023LLMCad: Fast and Scalable On-device Large Language Model Inferencecited by 45
  6. 2025Retrieval-Augmented Visual Question Answering via Built-in Autoregressive Search Enginescited by 13
  7. 2024TG-LLaVA: Text Guided LLaVA via Learnable Latent Embeddingscited by 8
  8. 2023VVS: Video-to-Video Retrieval with Irrelevant Frame Suppressioncited by 5
  9. 2024Towards Retrieval Augmented Generation over Large Video Librariescited by 4
  10. 2024Text-Video Multi-Grained Integration for Video Moment Montagecited by 3

Built on by

Not cited yet.