CLAICVAug 21, 2023

An Examination of the Compositionality of Large Generative Vision-Language Models

arXiv:2308.10509v235 citationsh-index: 14Has Code
Originality Synthesis-oriented
AI Analysis

This addresses the need for reliable evaluation in multimodal AI research, though it is incremental as it focuses on improving benchmarks rather than proposing new models or methods.

The paper tackles the problem of evaluating the compositionality of Generative Vision-Language Models (GVLMs) by identifying syntactical bias in current benchmarks and metrics, and it introduces a new benchmark called SADE to provide an unbiased assessment, with code and dataset made available.

With the success of Large Language Models (LLMs), many Generative Vision-Language Models (GVLMs) have been constructed via multimodal instruction tuning. However, the performance of GVLMs in multimodal compositional reasoning remains under-explored. In this paper, we examine both the evaluation metrics (VisualGPTScore, etc.) and current benchmarks for evaluating the compositionality of GVLMs. We identify the syntactical bias in current benchmarks, which is exploited by the linguistic capability of GVLMs. The bias renders VisualGPTScore an insufficient metric for assessing GVLMs. To combat this, we first introduce a SyntaxBias Score, leveraging LLMs to quantify such bias for mitigation. A challenging new task is subsequently added to evaluate the robustness of GVLMs against inherent inclination toward syntactical correctness. Using the bias-mitigated datasets and the new task, we propose a novel benchmark, namely SyntActically DE-biased benchmark (SADE). Our study provides an unbiased benchmark for the compositionality of GVLMs, facilitating future research in this direction (Code and dataset are available at https://github.com/TeleeMa/SADE).

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes