# Prototypicality Bias Reveals Blindspots in Multimodal Evaluation Metrics

arXiv: https://arxiv.org/abs/2601.04946 · HTML: https://mlsanity.com/paper/2601.04946/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 33 of its 63 references that are in the map.
- First posted on arXiv 8 Jan 2026.

## Builds on (most cited first, 30 of 33)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [CLIPScore: A Reference-free Evaluation Metric for Image Captioning](https://mlsanity.com/paper/2104.08718/) (2021)
- [A Survey on Bias and Fairness in Machine Learning](https://mlsanity.com/paper/1908.09635/) (2019)
- [InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models](https://mlsanity.com/paper/2504.10479/) (2025)
- [DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing](https://mlsanity.com/paper/2111.09543/) (2021)
- [MUSIQ: Multi-scale Image Quality Transformer](https://mlsanity.com/paper/2108.05997/) (2021)
- [C-Pack: Packed Resources For General Chinese Embeddings](https://mlsanity.com/paper/2309.07597/) (2023)
- [Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality](https://mlsanity.com/paper/2204.03162/) (2022)
- [Evaluating Text-to-Visual Generation with Image-to-Text Generation](https://mlsanity.com/paper/2404.01291/) (2024)
- [TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering](https://mlsanity.com/paper/2303.11897/) (2023)
- [SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformers](https://mlsanity.com/paper/2410.10629/) (2024)
- [AlignScore: Evaluating Factual Consistency with a Unified Alignment Function](https://mlsanity.com/paper/2305.16739/) (2023)
- [Ties Matter: Meta-Evaluating Modern Metrics with Pairwise Accuracy and Tie Calibration](https://mlsanity.com/paper/2305.14324/) (2023)
- [On the Blind Spots of Model-Based Evaluation Metrics for Text Generation](https://mlsanity.com/paper/2212.10020/) (2022)
- [Improving Pairwise Ranking for Multi-label Image Classification](https://mlsanity.com/paper/1704.03135/) (2017)
- [MM-SpuBench: Towards Better Understanding of Spurious Biases in Multimodal LLMs](https://mlsanity.com/paper/2406.17126/) (2024)
- [MLLM-Bench: Evaluating Multimodal LLMs with Per-sample Criteria](https://mlsanity.com/paper/2311.13951/) (2023)
- [textttModSCAN: Measuring Stereotypical Bias in Large Vision-Language Models from Vision and Language Modalities](https://mlsanity.com/paper/2410.06967/) (2024)
- [BIGbench: A Unified Benchmark for Evaluating Multi-dimensional Social Biases in Text-to-Image Models](https://mlsanity.com/paper/2407.15240/) (2024)
- [PrExMe! Large Scale Prompt Exploration of Open Source LLMs for Machine Translation and Summarization Evaluation](https://mlsanity.com/paper/2406.18528/) (2024)
- [AHELM: A Holistic Evaluation of Audio-Language Models](https://mlsanity.com/paper/2508.21376/) (2025)
- [Fooling the LVLM Judges: Visual Biases in LVLM-Based Evaluation](https://mlsanity.com/paper/2505.15249/) (2025)
- [PSL: Rethinking and Improving Softmax Loss from Pairwise Perspective for Recommendation](https://mlsanity.com/paper/2411.00163/) (2024)
- [Bias in Gender Bias Benchmarks: How Spurious Features Distort Evaluation](https://mlsanity.com/paper/2509.07596/) (2025)
- [FairPIVARA: Reducing and Assessing Biases in CLIP-Based Multimodal Models](https://mlsanity.com/paper/2409.19474/) (2024)
- [Deconstructing Bias: A Multifaceted Framework for Diagnosing Cultural and Compositional Inequities in Text-to-Image Generative Models](https://mlsanity.com/paper/2505.01430/) (2025)
- [Person-Centric Annotations of LAION-400M: Auditing Bias and Its Transfer to Models](https://mlsanity.com/paper/2510.03721/) (2025)
- [Exposing Blindspots: Cultural Bias Evaluation in Generative Image Models](https://mlsanity.com/paper/2510.20042/) (2025)
- [CROC: Evaluating and Training T2I Metrics with Pseudo- and Human-Labeled Contrastive Robustness Checks](https://mlsanity.com/paper/2505.11314/) (2025)
- [A Woman with a Knife or A Knife with a Woman? Measuring Directional Bias Amplification in Image Captions](https://mlsanity.com/paper/2503.07878/) (2025)

## Built on by (most cited first, 2 of 2)

- [Seeing Isn't Believing: Uncovering Blind Spots in Evaluator Vision-Language Models](https://mlsanity.com/paper/2604.21523/) (2026)
- [Do MLLM Judges Judge the Edit? Auditing Bias in Image Editing Evaluation with Verified Quality Preservation](https://mlsanity.com/paper/2610.01670/) (2026)

## Abstract (arXiv)

Automatic metrics are widely used to evaluate text-to-image models, often replacing human judgment in benchmarking, model selection, and large-scale data filtering. Yet they may reward images that look plausible or prototypical rather than images that faithfully satisfy the prompt. We identify prototypicality bias as a systematic blindspot in multimodal evaluation: metrics can prefer a semantically incorrect but visually or socially prototypical image over a correct but less prototypical one. We introduce PROTOBIAS, a controlled diagnostic benchmark across Animals, Objects, and Demography, where semantically correct images are contrasted with plausible prototypical adversaries containing a single controlled semantic violation. Grounded in prototype theory and social-category prototypicality, PROTOBIAS is constructed with multiple prompt generators, image generators, and independent VLM filters, and validated through prompt-quality, human-annotation, and image-quality controls. Using PROTOBIAS, we show that widely used embedding, reward, VQA-based, and VLM-as-judge metrics frequently fail these contrasts, while human judgments remain more faithful to semantic correctness. We further introduce PROTOSCORE, a lightweight contrastively trained evaluator, as an initial mitigation baseline. PROTOBIAS provides a focused benchmark for measuring prototypicality-driven metric failures and developing more semantically faithful T2I evaluators.