MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866909544011005952 |
|---|---|
| author | Qian, Yusu Ye, Hanrong Fauconnier, Jean-Philippe Grasch, Peter Yang, Yinfei Gan, Zhe |
| author_facet | Qian, Yusu Ye, Hanrong Fauconnier, Jean-Philippe Grasch, Peter Yang, Yinfei Gan, Zhe |
| contents | We introduce MIA-Bench, a new benchmark designed to evaluate multimodal large language models (MLLMs) on their ability to strictly adhere to complex instructions. Our benchmark comprises a diverse set of 400 image-prompt pairs, each crafted to challenge the models' compliance with layered instructions in generating accurate responses that satisfy specific requested patterns. Evaluation results from a wide array of state-of-the-art MLLMs reveal significant variations in performance, highlighting areas for improvement in instruction fidelity. Additionally, we create extra training data and explore supervised fine-tuning to enhance the models' ability to strictly follow instructions without compromising performance on other tasks. We hope this benchmark not only serves as a tool for measuring MLLM adherence to instructions, but also guides future developments in MLLM training methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2407_01509 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs Qian, Yusu Ye, Hanrong Fauconnier, Jean-Philippe Grasch, Peter Yang, Yinfei Gan, Zhe Computer Vision and Pattern Recognition Computation and Language We introduce MIA-Bench, a new benchmark designed to evaluate multimodal large language models (MLLMs) on their ability to strictly adhere to complex instructions. Our benchmark comprises a diverse set of 400 image-prompt pairs, each crafted to challenge the models' compliance with layered instructions in generating accurate responses that satisfy specific requested patterns. Evaluation results from a wide array of state-of-the-art MLLMs reveal significant variations in performance, highlighting areas for improvement in instruction fidelity. Additionally, we create extra training data and explore supervised fine-tuning to enhance the models' ability to strictly follow instructions without compromising performance on other tasks. We hope this benchmark not only serves as a tool for measuring MLLM adherence to instructions, but also guides future developments in MLLM training methods. |
| title | MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs |
| topic | Computer Vision and Pattern Recognition Computation and Language |
| url | https://arxiv.org/abs/2407.01509 |