Research map / Label-efficient and robust learning / Recognition and long-tail
Making Heads or Tails: Towards Semantically Consistent Visual Counterfactuals
Simon Vandenhende, Dhruv Mahajan, Filip Radenovic, Deepti Ghadiyaram
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 |
|---|---|---|---|
| add semantic consistency loss (ablation) | visual counterfactuals (VGG-16) · CUB | Same-KP: 8.3 -> 22.0 | Improved |
| add semantic consistency loss (ablation) | visual counterfactuals (VGG-16) · CUB | Near-KP: 54.6 -> 52.1 (-2.5) | Hurt |
| add semantic consistency loss (ablation) | visual counterfactuals (VGG-16) · CUB | #Edits: 5.5 -> 6.8 | Hurt |
| add multiple distractor images (ablation) | visual counterfactuals (VGG-16) · CUB | Near-KP: 54.6 -> 65.6 (+11.0) | Improved |
| add multiple distractor images (ablation) | visual counterfactuals (VGG-16) · CUB | runtime (s): 0.81 -> 9.98 | Hurt |
| add semantic loss on top of multiple distractors (ablation) | visual counterfactuals (VGG-16) · CUB | Same-KP: 13.8 -> 36.0 | Improved |
| add cell pre-filtering (ablation) | visual counterfactuals (VGG-16) · CUB | runtime (s): 10.82 -> 1.15 | Improved |
| add cell pre-filtering (ablation) | visual counterfactuals (VGG-16) · CUB | Same-KP: 36.0 -> 35.3 | No clear effect |
| use self-supervised DeepCluster instead of CUB classifier as auxiliary model (ablation) | semantic loss · CUB | Same-KP: 10.1 -> 22.0 | Improved |
| use ImageNet-1k supervised classifier instead of CUB classifier as auxiliary model (ablation) | semantic loss · CUB | Same-KP: 10.1 -> 19.3 | Improved |
| use SWAV instead of MoCo as auxiliary model (ablation) | semantic loss · CUB | Same-KP: 18.1 -> 22.1 | Improved |
| use our counterfactuals instead of Goyal et al. for human teaching with semantically-consistent pairs | machine teaching · CUB | test accuracy (%): 64.3 -> 82.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SCOUT — SCOUT: Self-Aware Discriminant Counterfactual Explanations (CVPR 2020) · 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
- Rethinking Visual Counterfactual Explanations Through Region Constraint (ICLR 2025 · similarity 0.91)
- Structure Your Data: Towards Semantic Graph Counterfactuals (ICML 2024 · similarity 0.90)
- Towards Fine-Grained Interpretability: Counterfactual Explanations for Misclassification with Saliency Partition (CVPR 2025 · similarity 0.90)
- V-CECE: Visual Counterfactual Explanations via Conceptual Edits (NeurIPS 2025 · similarity 0.89)
- Counterfactual Explanations for Misclassified Images: How Human and Machine Explanations Differ (Abstract Reprint) (AAAI 2024 · similarity 0.89)
- LeapFactual: Reliable Visual Counterfactual Explanation Using Conditional Flow Matching (NeurIPS 2025 · similarity 0.87)
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.
