# What makes a good metric? Evaluating automatic metrics for text-to-image consistency

arXiv: https://arxiv.org/abs/2412.13989 · HTML: https://mlsanity.com/paper/2412.13989/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- Builds on 50 of its 68 references that are in the map.
- First posted on arXiv 18 Dec 2024.

## Builds on (most cited first, 30 of 50)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [Hierarchical Text-Conditional Image Generation with CLIP Latents](https://mlsanity.com/paper/2204.06125/) (2022)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models](https://mlsanity.com/paper/2112.10741/) (2021)
- [Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering](https://mlsanity.com/paper/1612.00837/) (2016)
- [VQA: Visual Question Answering](https://mlsanity.com/paper/1505.00468/) (2015)
- [CLIPScore: A Reference-free Evaluation Metric for Image Captioning](https://mlsanity.com/paper/2104.08718/) (2021)
- [Shortcut Learning in Deep Neural Networks](https://mlsanity.com/paper/2004.07780/) (2020)
- [Making Pre-trained Language Models Better Few-shot Learners](https://mlsanity.com/paper/2012.15723/) (2020)
- [Energy and Policy Considerations for Deep Learning in NLP](https://mlsanity.com/paper/1906.02243/) (2019)
- [How Can We Know What Language Models Know?](https://mlsanity.com/paper/1911.12543/) (2019)
- [Annotation Artifacts in Natural Language Inference Data](https://mlsanity.com/paper/1803.02324/) (2018)
- [ImageNet-21K Pretraining for the Masses](https://mlsanity.com/paper/2104.10972/) (2021)
- [When and why vision-language models behave like bags-of-words, and what to do about it?](https://mlsanity.com/paper/2210.01936/) (2022)
- [Adversarial Examples for Evaluating Reading Comprehension Systems](https://mlsanity.com/paper/1707.07328/) (2017)
- [Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality](https://mlsanity.com/paper/2204.03162/) (2022)
- [A systematic study of the class imbalance problem in convolutional neural networks](https://mlsanity.com/paper/1710.05381/) (2017)
- [Large Language Models Are Not Robust Multiple Choice Selectors](https://mlsanity.com/paper/2309.03882/) (2023)
- [Dynabench: Rethinking Benchmarking in NLP](https://mlsanity.com/paper/2104.14337/) (2021)
- [TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering](https://mlsanity.com/paper/2303.11897/) (2023)
- [AI and the Everything in the Whole Wide World Benchmark](https://mlsanity.com/paper/2111.15366/) (2021)
- [Don't Just Assume; Look and Answer: Overcoming Priors for Visual Question Answering](https://mlsanity.com/paper/1712.00377/) (2017)
- [Measurement and Fairness](https://mlsanity.com/paper/1912.05511/) (2019)
- [Masked Language Modeling and the Distributional Hypothesis: Order Word Matters Pre-training for Little](https://mlsanity.com/paper/2104.06644/) (2021)
- [Davidsonian Scene Graph: Improving Reliability in Fine-grained Evaluation for Text-to-Image Generation](https://mlsanity.com/paper/2310.18235/) (2023)
- [Scaling Autoregressive Multi-Modal Models: Pretraining and Instruction Tuning](https://mlsanity.com/paper/2309.02591/) (2023)
- [An Empirical Study on Robustness to Spurious Correlations using Pre-trained Language Models](https://mlsanity.com/paper/2007.06778/) (2020)
- [Yin and Yang: Balancing and Answering Binary Visual Questions](https://mlsanity.com/paper/1511.05099/) (2015)
- [Towards Causal VQA: Revealing and Reducing Spurious Correlations by Invariant and Covariant Semantic Editing](https://mlsanity.com/paper/1912.07538/) (2019)

## Built on by (most cited first, 12 of 12)

- [Maestro: Self-Improving Text-to-Image Generation via Agent Orchestration](https://mlsanity.com/paper/2509.10704/) (2025)
- [Calibrating LLM Judges: Linear Probes for Fast and Reliable Uncertainty Estimation](https://mlsanity.com/paper/2512.22245/) (2025)
- [DIMCIM: A Quantitative Evaluation Framework for Default-mode Diversity and Generalization in Text-to-Image Generative Models](https://mlsanity.com/paper/2506.05108/) (2025)
- [CORDIAL: Can Multimodal Large Language Models Effectively Understand Coherence Relationships?](https://mlsanity.com/paper/2502.11300/) (2025)
- [Multimodal Language Models as Text-to-Image Model Evaluators](https://mlsanity.com/paper/2505.00759/) (2025)
- [SANEval: Open-Vocabulary Compositional Benchmarks with Failure-mode Diagnosis](https://mlsanity.com/paper/2602.00249/) (2026)
- [Do Image-Text Metrics Respect Semantic Invariances?](https://mlsanity.com/paper/2605.24702/) (2026)
- [Physics Question Scene Graph: Fine-grained Evaluation of Physical Plausibility in Text-to-Video Generation](https://mlsanity.com/paper/2606.25306/) (2026)
- [Cost Savings from Automatic Quality Assessment of Generated Images](https://mlsanity.com/paper/2510.16179/) (2025)
- [System-Mediated Attention Imbalances Make Vision-Language Models Say Yes](https://mlsanity.com/paper/2601.12430/) (2026)
- [Frozen Scenes, Shifting Winners: Configuration Fragility in Text-to-3D Evaluation](https://mlsanity.com/paper/2610.00447/) (2026)
- [VisExMEM: Visual Evidence-Based Explainable Multimodal Evaluation Metric](https://mlsanity.com/paper/2610.16441/) (2026)

## Abstract (arXiv)

Language models are increasingly being incorporated as components in larger AI systems for various purposes, from prompt optimization to automatic evaluation. In this work, we analyze the construct validity of four recent, commonly used methods for measuring text-to-image consistency - CLIPScore, TIFA, VPEval, and DSG - which rely on language models and/or VQA models as components. We define construct validity for text-image consistency metrics as a set of desiderata that text-image consistency metrics should have, and find that no tested metric satisfies all of them. We find that metrics lack sufficient sensitivity to language and visual properties. Next, we find that TIFA, VPEval and DSG contribute novel information above and beyond CLIPScore, but also that they correlate highly with each other. We also ablate different aspects of the text-image consistency metrics and find that not all model components are strictly necessary, also a symptom of insufficient sensitivity to visual information. Finally, we show that all three VQA-based metrics likely rely on familiar text shortcuts (such as yes-bias in QA) that call their aptitude as quantitative evaluations of model performance into question.