Elamur Research⌕
☰
home / papers / 2023 / LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models

LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models

Paper · 2023 · Yanwei Li, Chengyao Wang and Jiaya Jia

Summary

Vision Language Models handle single images well but struggle with long videos. Each frame normally needs dozens to hundreds of visual tokens, so hour-long footage quickly exceeds what an LLM can process. LLaMA-VID, from CUHK and SmartMore, compresses every frame into just two tokens. A text decoder (BERT or QFormer) turns the user's instruction into queries. A "context attention" module uses these queries to collect the visual features most relevant to the question and condenses them into one context token. In parallel, average pooling of the frame's visual embedding produces a content token. The content token can be shrunk to a single token for video or kept at full resolution for still images. Simple linear projectors map both tokens into the LLM's embedding space. The authors also built an instruction dataset of 9K movie-level conversations for long-video reasoning. Using 7B LLMs, the model outperforms prior methods on most image and video benchmarks, which makes hour-long video understanding practical.

Key points

  • Token budget: BLIP-style models use 32 tokens per image and LLaVA over 256. At that rate, a 10K-frame video would need more than 320K tokens. LLaMA-VID needs only 2 tokens per video frame.
  • Context token: text-guided queries attend over ViT patch features (patch size 14), and the result is averaged into one instruction-aware token.
  • Content token: adaptive average pooling yields n tokens, where n ranges from 1 to N patches. This trades detail against compute.
  • New dataset: an instruction dataset of 9K movie-level conversations covers plot reasoning and detail understanding in hour-long videos.
  • Efficiency and results: training finishes in about 2 days on one machine with 8×A100 GPUs. With 7B LLMs such as Vicuna-7B, the model leads on most benchmarks reported, including MSVD, MSRVTT, ActivityNet, GQA, MME and POPE.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Chinese University of Hong KongHong Kong · China159 works · 16,154 citations

 Fields, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksComputer Vision and Pattern Recognition · 3,486 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,894 works

Cite

Cite as Yanwei Li, Chengyao Wang and Jiaya Jia. 2023. LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models. arXiv:2311.17043.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2311.17043 · arXiv 2311.17043

@misc{li2023llama,
  title = {{LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models}},
  author = {Yanwei Li, Chengyao Wang and Jiaya Jia},
  year = {2023},
  doi = {10.48550/arxiv.2311.17043},
  eprint = {2311.17043},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2311.17043}
}

References

65 references, 45 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Sharegpt. https://sharegpt.com/, 2023. ↗
  2. Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katherine Millican, Malcolm Reynolds, et al. Flamingo: a visual language model for few-shot learning. In NeurIPS, 2022.
  3. Anthropic. Claude 2. https://www.anthropic.com/index/claude-2, 2023. ↗
  4. Jinze Bai, Shuai Bai, Shusheng Yang, Shijie Wang, Sinan Tan, Peng Wang, Junyang Lin, Chang Zhou, and Jingren Zhou. Qwen-vl: A frontier large vision-language model with versatile abilities. arXiv:2308.12966, 2023.
  5. Max Bain, Arsha Nagrani, Gül Varol, and Andrew Zisserman. Frozen in time: A joint video and image encoder for end-to-end retrieval. In ICCV, 2021. find
  6. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. In NeurIPS, 2020.
  7. Fabian Caba Heilbron, Victor Escorcia, Bernard Ghanem, and Juan Carlos Niebles. Activitynet: A large-scale video benchmark for human activity understanding. In CVPR, 2015. find
  8. David Chen and William B Dolan. Collecting highly parallel data for paraphrase evaluation. In ACL, 2011.
  9. Keqin Chen, Zhao Zhang, Weili Zeng, Richong Zhang, Feng Zhu, and Rui Zhao. Shikra: Unleashing multimodal llm’s referential dialogue magic. arXiv:2306.15195, 2023.
  10. Shouyuan Chen, Sherman Wong, Liangjian Chen, and Yuandong Tian. Extending context window of large language models via positional interpolation. arXiv:2306.15595, 2023. find
  11. Xinlei Chen, Hao Fang, Tsung-Yi Lin, Ramakrishna Vedantam, Saurabh Gupta, Piotr Dollár, and C Lawrence Zitnick. Microsoft coco captions: Data collection and evaluation server. arXiv:1504.00325, 2015.
  12. Yukang Chen, Shengju Qian, Haotian Tang, Xin Lai, Zhijian Liu, Song Han, and Jiaya Jia. Longlora: Efficient fine-tuning of long-context large language models. arXiv:2309.12307, 2023.
  13. Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E. Gonzalez, Ion Stoica, and Eric P. Xing. Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality. https://lmsys.org/blog/2023-03-30-vicuna/, 2023. ↗
  14. Wenliang Dai, Junnan Li, Dongxu Li, Anthony Meng Huat Tiong, Junqi Zhao, Weisheng Wang, Boyang Li, Pascale Fung, and Steven Hoi. Instructblip: Towards general-purpose vision-language models with instruction tuning. arXiv:2305.06500, 2023.
  15. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv:1810.04805, 2018.
  16. Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. In ICLR, 2021.
  17. Yuxin Fang, Wen Wang, Binhui Xie, Quan Sun, Ledell Wu, Xinggang Wang, Tiejun Huang, Xinlong Wang, and Yue Cao. Eva: Exploring the limits of masked visual representation learning at scale. In CVPR, 2023. find
  18. Chaoyou Fu, Peixian Chen, Yunhang Shen, Yulei Qin, Mengdan Zhang, Xu Lin, Zhenyu Qiu, Wei Lin, Jinrui Yang, Xiawu Zheng, et al. Mme: A comprehensive evaluation benchmark for multimodal large language models. arXiv:2306.13394, 2023.
  19. Yash Goyal, Tejas Khot, Douglas Summers-Stay, Dhruv Batra, and Devi Parikh. Making the v in vqa matter: Elevating the role of image understanding in visual question answering. In CVPR, 2017.
  20. Danna Gurari, Qing Li, Abigale J Stangl, Anhong Guo, Chi Lin, Kristen Grauman, Jiebo Luo, and Jeffrey P Bigham. Vizwiz grand challenge: Answering visual questions from blind people. In CVPR, 2018. find
  21. Qingqiu Huang, Yu Xiong, Anyi Rao, Jiaze Wang, and Dahua Lin. Movienet: A holistic dataset for movie understanding. In ECCV, 2020. find
  22. Drew A Hudson and Christopher D Manning. Gqa: A new dataset for real-world visual reasoning and compositional question answering. In CVPR, 2019.
  23. IDEFICS. Introducing idefics: An open reproduction of state-of-the-art visual language model. https://huggingface.co/blog/idefics, 2023. ↗
  24. Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc Le, Yun-Hsuan Sung, Zhen Li, and Tom Duerig. Scaling up visual and vision-language representation learning with noisy text supervision. In ICML, 2021.
  25. Sahar Kazemzadeh, Vicente Ordonez, Mark Matten, and Tamara Berg. Referitgame: Referring to objects in photographs of natural scenes. In EMNLP, 2014.
  26. Ranjay Krishna, Yuke Zhu, Oliver Groth, Justin Johnson, Kenji Hata, Joshua Kravitz, Stephanie Chen, Yannis Kalantidis, Li-Jia Li, David A Shamma, et al. Visual genome: Connecting language and vision using crowdsourced dense image annotations. IJCV, 2017.
  27. Xin Lai, Zhuotao Tian, Yukang Chen, Yanwei Li, Yuhui Yuan, Shu Liu, and Jiaya Jia. Lisa: Reasoning segmentation via large language model. arXiv:2308.00692, 2023. find
  28. Bohao Li, Rui Wang, Guangzhi Wang, Yuying Ge, Yixiao Ge, and Ying Shan. Seed-bench: Benchmarking multimodal llms with generative comprehension. arXiv:2307.16125, 2023.
  29. Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. arXiv:2301.12597, 2023.
  30. KunChang Li, Yinan He, Yi Wang, Yizhuo Li, Wenhai Wang, Ping Luo, Yali Wang, Limin Wang, and Yu Qiao. Videochat: Chat-centric video understanding. arXiv:2305.06355, 2023.
  31. Yifan Li, Yifan Du, Kun Zhou, Jinpeng Wang, Wayne Xin Zhao, and Ji-Rong Wen. Evaluating object hallucination in large vision-language models. arXiv:2305.10355, 2023.
  32. Haotian Liu, Chunyuan Li, Yuheng Li, and Yong Jae Lee. Improved baselines with visual instruction tuning. arXiv:2310.03744, 2023.
  33. Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. In NeruIPS, 2023.
  34. Ruyang Liu, Chen Li, Yixiao Ge, Ying Shan, Thomas H Li, and Ge Li. One for all: Video conversation is feasible without video instruction tuning. arXiv:2309.15785, 2023. find
  35. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. Roberta: A robustly optimized bert pretraining approach. arXiv:1907.11692, 2019.
  36. Yuan Liu, Haodong Duan, Yuanhan Zhang, Bo Li, Songyang Zhang, Wangbo Zhao, Yike Yuan, Jiaqi Wang, Conghui He, Ziwei Liu, et al. Mmbench: Is your multi-modal model an all-around player? arXiv:2307.06281, 2023.
  37. Pan Lu, Swaroop Mishra, Tanglin Xia, Liang Qiu, Kai-Wei Chang, Song-Chun Zhu, Oyvind Tafjord, Peter Clark, and Ashwin Kalyan. Learn to explain: Multimodal reasoning via thought chains for science question answering. In NeurIPS, 2022. find
  38. Ruipu Luo, Ziwang Zhao, Min Yang, Junwei Dong, Minghui Qiu, Pengcheng Lu, Tao Wang, and Zhongyu Wei. Valley: Video assistant with large language model enhanced ability. arXiv:2306.07207, 2023. — [arXiv:2306.07207](https://arxiv.org/abs/2306.07207) find
  39. Muhammad Maaz, Hanoona Rasheed, Salman Khan, and Fahad Shahbaz Khan. Video-chatgpt: Towards detailed video understanding via large vision and language models. arXiv:2306.05424, 2023.
  40. Junhua Mao, Jonathan Huang, Alexander Toshev, Oana Camburu, Alan L Yuille, and Kevin Murphy. Generation and comprehension of unambiguous object descriptions. In CVPR, 2016.
  41. Junhua Mao, Jonathan Huang, Alexander Toshev, Oana Camburu, Alan L Yuille, and Kevin Murphy. Generation and comprehension of unambiguous object descriptions. In CVPR, 2016.
  42. Anand Mishra, Shashank Shekhar, Ajeet Kumar Singh, and Anirban Chakraborty. Ocr-vqa: Visual question answering by reading text in images. In ICDAR, 2019. find
  43. OpenAI. Chatgpt. https://openai.com/blog/chatgpt/, 2023. ↗
  44. OpenAI. Gpt-4 technical report. arXiv:2303.08774, 2023.
  45. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback. In NeurIPS, 2022.
  46. Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In ICML, 2021.
  47. Dustin Schwenk, Apoorv Khandelwal, Christopher Clark, Kenneth Marino, and Roozbeh Mottaghi. A-okvqa: A benchmark for visual question answering using world knowledge. In ECCV, 2022.
  48. Piyush Sharma, Nan Ding, Sebastian Goodman, and Radu Soricut. Conceptual captions: A cleaned, hypernymed, image alt-text dataset for automatic image captioning. In ACL, 2018.
  49. Oleksii Sidorov, Ronghang Hu, Marcus Rohrbach, and Amanpreet Singh. Textcaps: a dataset for image captioning with reading comprehension. In ECCV, 2020.
  50. Amanpreet Singh, Vivek Natarajan, Meet Shah, Yu Jiang, Xinlei Chen, Dhruv Batra, Devi Parikh, and Marcus Rohrbach. Towards vqa models that can read. In CVPR, 2019. find
  51. Jianlin Su, Yu Lu, Shengfeng Pan, Ahmed Murtadha, Bo Wen, and Yunfeng Liu. Roformer: Enhanced transformer with rotary position embedding. arXiv:2104.09864, 2021.
  52. Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. Stanford alpaca: An instruction-following llama model. https://github.com/tatsu-lab/ stanford_alpaca, 2023. ↗
  53. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. Llama: Open and efficient foundation language models. arXiv:2302.1…
  54. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. Llama 2: Open foundation and fine-tuned chat models. arXiv:2307.09288, 2023.
  55. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. In NeurIPS, 2017.
  56. Jason Wei, Maarten Bosma, Vincent Y Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. Finetuned language models are zero-shot learners. arXiv:2109.01652, 2021.
  57. Chenfei Wu, Shengming Yin, Weizhen Qi, Xiaodong Wang, Zecheng Tang, and Nan Duan. Visual chatgpt: Talking, drawing and editing with visual foundation models. arXiv:2303.04671, 2023.
  58. Jun Xu, Tao Mei, Ting Yao, and Yong Rui. Msr-vtt: A large video description dataset for bridging video and language. In CVPR, 2016. find
  59. Antoine Yang, Antoine Miech, Josef Sivic, Ivan Laptev, and Cordelia Schmid. Zero-shot video question answering via frozen bidirectional language models. In NeurIPS, 2022. find
  60. Rui Yang, Lin Song, Yanwei Li, Sijie Zhao, Yixiao Ge, Xiu Li, and Ying Shan. Gpt4tools: Teaching large language model to use tools via self-instruction. arXiv:2305.18752, 2023.
  61. Zhengyuan Yang, Linjie Li, Kevin Lin, Jianfeng Wang, Chung-Ching Lin, Zicheng Liu, and Lijuan Wang. The dawn of lmms: Preliminary explorations with gpt-4v (ision). arXiv:2309.17421, 2023.
  62. Hang Zhang, Xin Li, and Lidong Bing. Video-llama: An instruction-tuned audio-visual language model for video understanding. arXiv:2306.02858, 2023.
  63. Renrui Zhang, Jiaming Han, Aojun Zhou, Xiangfei Hu, Shilin Yan, Pan Lu, Hongsheng Li, Peng Gao, and Yu Qiao. Llama-adapter: Efficient fine-tuning of language models with zero-init attention. arXiv:2303.16199, 2023.
  64. Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, et al. Opt: Open pre-trained transformer language models. arXiv:2205.01068, 2022.
  65. Deyao Zhu, Jun Chen, Xiaoqian Shen, Xiang Li, and Mohamed Elhoseiny. Minigpt-4: Enhancing vision-language understanding with advanced large language models. arXiv:2304.10592, 2023.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/li-2023-llama-vid-image-worth-2/