Scalable Performance Analysis for Vision-Language Models

Proceedings of the The 12th Joint Conference on Lexical and Computational Semantics (*SEM 2023), pages 284–294 July 13-14, 2023 ©2023 Association for Computational Linguistics Scalable Performance Analysis for Vision-Language Models Santiago Castro∗ Oana Ignat∗ Rada Mihalcea University of Michigan – Ann Arbor, USA {sacastro,oignat,mihalcea}@umich.edu Abstract Joint vision-language models have shown great performance over a diverse set of tasks. However, little is known about their limitations, as the high dimensional space learned by these models makes it difficult to identify semantic errors. Recent work has addressed this problem by designing highly controlled probing task benchmarks. Our paper introduces a more scalable solution that relies on already annotated benchmarks. Our method consists of extracting a large set of diverse features from a vision-language benchmark and measuring their correlation with the output of the target model. We confirm previous findings that CLIP behaves like a bag of words model and performs better with nouns and verbs; we also uncover novel insights such as CLIP getting confused by concrete words. Our framework is available at https://github.com/MichiganNLP/ Scalable-VLM-Probing and can be used with other multimodal models and benchmarks. 1 Introduction Recent years have witnessed an explosion of vision- language models (Lu et al., 2019; Li et al., 2019; Zhang et al., 2021; Radford et al., 2021; Singh et al., 2022). These models have shown great per- formance in a variety of tasks, such as image/video classification and text-image/video retrieval (Rad- ford et al., 2021; Luo et al., 2022), even without leveraging task-specific or in-domain training. In addition, these models have shown to be practi- cal when leveraged as underlying models for text- to-image generation such as DALL-E 2 (Ramesh et al., 2022) and image captioning such as Clip- Cap (Mokady et al., 2021). Little is however known about the limita- tions of these models. Recent work, such *Equal contribution. Image Caption: Girl is standing in the grass. Pos Neg CLIP Features of original word (girl), replacement word (dog) & words in common (stand, grass) Neg score (N) Pos score (P) Correlation CLIP P - N = score diff. (D) FEMALE P +0.022 ANIMAL N -0.005 PLANT D +0.022 Feature Score Figure 1: We propose a simple framework to analyze CLIP performance on SVO-Probes data. We test CLIP on the benchmark, extract a diverse set of semantic fea- tures from the data, and measure the correlation between each feature and the CLIP score (P, N, or D). Features with positive correlation (e.g., Female, Plant) impact positively the model performance, while features with negative correlation (e.g., Animal) impact negatively the model performance. as Winoground (Thrush et al., 2022), SVO- Probes (Hendricks and Nematzadeh, 2021), or VALSE (Parcalabescu et al., 2022), have designed benchmark probing tasks by annotating data to fol- low specific properties (i.e., object color, location, size, swapping word order, replacing words). This line of research led to valuable insights into the limitations of current state-of-the-art multi-modal models such as CLIP (Radford et al., 2021) and ViLBERT (Lu et al., 2019). An important limitation of current …

Content truncated to 500 words