Research map / Label-efficient and robust learning / Recognition and long-tail
Robustness to Spurious Correlations in Text Classification via Automatically Generated Counterfactuals
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Augment training with automatically generated counterfactuals using predicted causal terms | LogisticRegression · IMDB-L counterfactual test | accuracy: .615 -> .744 | Helped |
| Augment training with automatically generated counterfactuals using predicted causal terms | LogisticRegression · IMDB-S counterfactual test | accuracy: .605 -> .660 | Helped |
| Augment training with automatically generated counterfactuals using predicted causal terms | LogisticRegression · Kindle counterfactual test | accuracy: .514 -> .624 | Helped |
| Augment with counterfactuals generated from human-annotated causal terms across the whole vocabulary | LogisticRegression · IMDB-L counterfactual test | accuracy: .615 -> .857 | Helped |
| Augment with counterfactuals generated from human-annotated causal terms across the whole vocabulary | LogisticRegression · IMDB-S counterfactual test | accuracy: .605 -> .726 | Helped |
| Augment with counterfactuals generated from human-annotated causal terms across the whole vocabulary | LogisticRegression · Kindle counterfactual test | accuracy: .514 -> .720 | Helped |
| Add counterfactual augmentation using predicted causal terms | LogisticRegression · IMDB-L original test | accuracy: .816 -> .742 | Hurt |
| Add counterfactual augmentation using predicted causal terms | LogisticRegression · IMDB-S original test | accuracy: .711 -> .685 | Hurt |
| Add counterfactual augmentation using predicted causal terms | LogisticRegression · Kindle original test | accuracy: .888 -> .866 | Hurt |
| Replace predicted causal terms with human-annotated terms from the whole vocabulary (ablation) | Counterfactual augmentation · IMDB-L counterfactual test | accuracy: .744 -> .857 | Helped |
| Replace predicted causal terms with human-annotated terms from the whole vocabulary (ablation) | Counterfactual augmentation · IMDB-S counterfactual test | accuracy: .660 -> .726 | Helped |
| Replace predicted causal terms with human-annotated terms from the whole vocabulary (ablation) | Counterfactual augmentation · Kindle counterfactual test | accuracy: .624 -> .720 | Helped |
Closest papers by meaning
- Counterfactual Invariance to Spurious Correlations in Text Classification (NeurIPS 2021 · similarity 0.91)
- Learning The Difference That Makes A Difference With Counterfactually-Augmented Data (ICLR 2020 · similarity 0.90)
- Promoting Counterfactual Robustness through Diversity (AAAI 2024 · similarity 0.85)
- A General Search-Based Framework for Generating Textual Counterfactual Explanations (AAAI 2024 · similarity 0.85)
- Iterative Counterfactual Data Augmentation (AAAI 2025 · similarity 0.84)
- Conditional Counterfactual Mean Embeddings: Doubly Robust Estimation and Learning Rates (NeurIPS 2026 · similarity 0.84)
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.