AIM: Adaptive Inference of Multi-Modal LLMs via Token Merging and Pruning

• URL: http://arxiv.org/abs/2412.03248v1
• Date: Wed, 04 Dec 2024 11:47:57 GMT
• Title: AIM: Adaptive Inference of Multi-Modal LLMs via Token Merging and Pruning
• Authors: Yiwu Zhong, Zhuoming Liu, Yin Li, Liwei Wang,
• Abstract summary: We propose a training-free adaptive inference method for multi-modal LLMs. With a minimalist design, our method can be applied to both video and image LLMs. Under a similar computational cost, our method outperforms the state-of-the-art methods in long video understanding.
• Score: 19.68349294206012
• License:
• Abstract: Large language models (LLMs) have enabled the creation of multi-modal LLMs that exhibit strong comprehension of visual data such as images and videos. However, these models usually rely on extensive visual tokens from visual encoders, leading to high computational demands, which limits their applicability in resource-constrained environments and for long-context tasks. In this work, we propose a training-free adaptive inference method for multi-modal LLMs that can accommodate a broad range of efficiency requirements with a minimum performance drop. Our method consists of a) iterative token merging based on embedding similarity before LLMs, and b) progressive token pruning within LLM layers based on multi-modal importance. With a minimalist design, our method can be applied to both video and image LLMs. Extensive experiments on diverse video and image benchmarks demonstrate that, our method substantially reduces computation load (e.g., a $\textbf{7-fold}$ reduction in FLOPs) while preserving the performance of video and image LLMs. Further, under a similar computational cost, our method outperforms the state-of-the-art methods in long video understanding (e.g., $\textbf{+4.6}$ on MLVU). Additionally, our in-depth analysis provides insights into token redundancy and LLM layer behaviors, offering guidance for future research in designing efficient multi-modal LLMs. Our code will be available at https://github.com/LaVi-Lab/AIM.

Related papers

• OLA-VLM: Elevating Visual Perception in Multimodal LLMs with Auxiliary Embedding Distillation [95.78870389271832]
  The standard practice for developing contemporary MLLMs is to feed features from vision encoder(s) into the LLM and train with natural language supervision. We propose OLA-VLM, the first approach distilling knowledge into the LLM's hidden representations from a set of target visual representations. We show that OLA-VLM boosts performance by an average margin of up to 2.5% on various benchmarks, with a notable improvement of 8.7% on the Depth task in CV-Bench.
  arXiv  Detail & Related papers  (2024-12-12T18:55:18Z)
• TS-LLaVA: Constructing Visual Tokens through Thumbnail-and-Sampling for Training-Free Video Large Language Models [52.590072198551944]
  Recent advances in multimodal Large Language Models (LLMs) have shown great success in understanding multi-modal contents. For video understanding tasks, training-based video LLMs are difficult to build due to the scarcity of high-quality, curated video-text paired data. In this work, we explore the limitations of the existing compression strategies for building a training-free video LLM.
  arXiv  Detail & Related papers  (2024-11-17T13:08:29Z)
• LLaVA-KD: A Framework of Distilling Multimodal Large Language Models [70.19607283302712]
  We propose a novel framework to transfer knowledge from l-MLLM to s-MLLM. Specifically, we introduce Multimodal Distillation (MDist) to minimize the divergence between the visual-textual output distributions of l-MLLM and s-MLLM. We also propose a three-stage training scheme to fully exploit the potential of s-MLLM.
  arXiv  Detail & Related papers  (2024-10-21T17:41:28Z)
• SWIFT: On-the-Fly Self-Speculative Decoding for LLM Inference Acceleration [10.970637831760136]
  Speculative decoding (SD) has emerged as a widely used paradigm to accelerate the inference of large language models (LLMs) We introduce SWIFT, an on-the-fly self-speculative decoding algorithm that adaptively selects intermediate layers of LLMs to skip during inference. We show that SWIFT can achieve over a 1.3x-1.6x speedup while preserving the original distribution of the generated text.
  arXiv  Detail & Related papers  (2024-10-09T14:15:30Z)
• Dense Connector for MLLMs [89.50595155217108]
  We introduce the Dense Connector - a plug-and-play vision-language connector that significantly enhances existing MLLMs. Building on this, we also propose the Efficient Dense Connector, which achieves performance comparable to LLaVA-v1.5 with only 25% of the visual tokens. Our model, trained solely on images, showcases remarkable zero-shot capabilities in video understanding as well.
  arXiv  Detail & Related papers  (2024-05-22T16:25:03Z)
• From Image to Video, what do we need in multimodal LLMs? [19.85928004619801]
  Multimodal Large Language Models (MLLMs) have demonstrated profound capabilities in understanding multimodal information. We propose RED-VILLM, a Resource-Efficient Development pipeline for Video LLMs from Image LLMs. Our approach highlights the potential for a more cost-effective and scalable advancement in multimodal models.
  arXiv  Detail & Related papers  (2024-04-18T02:43:37Z)
• ST-LLM: Large Language Models Are Effective Temporal Learners [58.79456373423189]
  Large Language Models (LLMs) have showcased impressive capabilities in text comprehension and generation. How to effectively encode and understand videos in video-based dialogue systems remains to be solved. We propose ST-LLM, an effective video-LLM baseline with spatial-temporal sequence modeling inside LLM.
  arXiv  Detail & Related papers  (2024-03-30T10:11:26Z)
• CLAMP: Contrastive LAnguage Model Prompt-tuning [89.96914454453791]
  We show that large language models can achieve good image classification performance when adapted this way. Our approach beats state-of-the-art mLLMs by 13% and slightly outperforms contrastive learning with a custom text model.
  arXiv  Detail & Related papers  (2023-12-04T05:13:59Z)
• InfMLLM: A Unified Framework for Visual-Language Tasks [44.29407348046122]
  multimodal large language models (MLLMs) have attracted growing interest. This work delves into enabling LLMs to tackle more vision-language-related tasks. InfMLLM achieves either state-of-the-art (SOTA) performance or performance comparable to recent MLLMs.
  arXiv  Detail & Related papers  (2023-11-12T09:58:16Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.