Summary of Milebench: Benchmarking Mllms in Long Context, by Dingjie Song et al.
MileBench: Benchmarking MLLMs in Long Context
by Dingjie Song, Shunian Chen, Guiming Hardy Chen, Fei Yu, Xiang Wan, Benyou Wang
First submitted to arxiv on: 29 Apr 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper addresses the limitations of existing benchmarks for Multimodal Large Language Models (MLLMs) by introducing a new benchmark called MileBench. The existing benchmarks often focus on single-image and short-text samples, which may not accurately reflect the performance challenges of MLLMs in real-world scenarios. MileBench is designed to test the MultImodal Long-contExt capabilities of MLLMs, comprising multimodal long contexts and multiple tasks requiring comprehension and generation. The benchmark includes two distinct evaluation sets: diagnostic and realistic, to assess the models’ long-context adaptation capacity and ability to complete tasks in long-context scenarios. Experimental results show that while closed-source GPT-4o outperforms others, most open-source MLLMs struggle in long-context situations, with a widening performance gap as the number of images increases. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper introduces MileBench, a new benchmark for Multimodal Large Language Models (MLLMs), to test their ability to handle long contexts and multiple images. The existing benchmarks are limited to single-image and short-text samples, which doesn’t reflect real-world scenarios. MileBench is designed to fill this gap by including multimodal long contexts and multiple tasks requiring comprehension and generation. The benchmark also includes two evaluation sets: diagnostic and realistic. |
Keywords
» Artificial intelligence » Gpt
