Research map / Vision-language and multimodal / Human and embodied modeling
Domino: Discovering Systematic Errors with Cross-Modal Embeddings
Sabri Eyuboglu, Maya Varma, Khaled Kamal Saab, Jean-Benoit Delbrouck, Christopher Lee-Messer, Jared Dunnmon, James Zou, Christopher Re
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use cross-modal CLIP embeddings instead of BiT embeddings (ablation) | Domino on natural images with synthetic models · Natural images | mean precision-at-10: a 9 percentage point increase over BiT embeddings | Improved |
| Use cross-modal CLIP embeddings instead of random activations (ablation) | Domino on natural images with synthetic models · Natural images | mean precision-at-10: a 23 percentage point increase over random activations | Improved |
| Use CLIP embeddings instead of BiT embeddings (ablation) | Domino on natural images with trained models · Natural images | mean precision-at-10 | No clear effect |
| Use CLIP embeddings instead of trained-classifier activations (ablation) | Domino on natural images with trained models · Natural images | mean precision-at-10: nearly 15 percentage points | Improved |
| Use ConVIRT embeddings instead of BiT embeddings (ablation) | Domino on medical images with synthetic models · Medical images | mean precision-at-10: a 7 point improvement over the best unimodal embeddings (BiT) | Improved |
| Use cross-modal embeddings instead of unimodal embeddings (ablation) | Domino on medical time-series with synthetic models · Medical time-series | mean precision-at-10: a 17 point gain over using unimodal embeddings 0.532 | Improved |
| Use the error-aware mixture model instead of George (ablation) | Domino with cross-modal embeddings on natural-image noisy and rare slices · Natural images, noisy and rare slices | mean precision-at-10: a 105% improvement over the next-best method, George | Improved |
| Use the error-aware mixture model instead of ConfusionSDM (ablation) | Domino with cross-modal embeddings on correlation slices · Natural images, correlation slices | slice recovery performance | Hurt |
| Generate natural-language slice descriptions from cross-modal embeddings | Domino on natural-image slice settings · Natural images, slice descriptions | settings with exact slice name ranked first: 34.7% of rare slices, 41.0% of correlation slices, and 39.0% of noisy label slices | Improved |
Later papers that compared with this method
1 later accepted paper reports Domino as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Discover and Mitigate Multiple Biased Subgroups in Image Classifiers (CVPR 2024 · better on 4 of 4 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- CB-SLICE: Concept-Based Interpretable Error Slice Discovery (ICML 2026 · similarity 0.88)
- Enhancing vision-language models for medical imaging: bridging the 3D gap with innovative slice selection (NeurIPS 2024 · similarity 0.84)
- HiBug2: Efficient and Interpretable Error Slice Discovery for Comprehensive Model Debugging (ICLR 2025 · similarity 0.84)
- SliceGPT: Compress Large Language Models by Deleting Rows and Columns (ICLR 2024 · similarity 0.83)
- SANEval: Open-Vocabulary Compositional Benchmarks with Failure-mode Diagnosis (NeurIPS 2026 · similarity 0.83)
- SASA: Subspace-Aware Sparse Autoencoders for Effective Mechanistic Interpretability (NeurIPS 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
