Research map / Probabilistic and scientific modeling / Causal discovery
Debiasing NLU Models via Causal Intervention and Counterfactual Reasoning
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 |
|---|---|---|---|
| Use CICR-SUM instead of BERT | BERT · SNLI hard | accuracy: +5.3 points | Helped |
| Use CICR-FC instead of BERT | BERT · SNLI hard | accuracy: +4.11 points | Helped |
| Use CICR-FC instead of BERT | BERT · FEVER symmetric test v1 | accuracy: +13.52 points | Helped |
| Use CICR-SUM instead of BERT | BERT · FEVER symmetric test v1 | accuracy: +14.95 points | Helped |
| Use CICR-FC instead of BERT | BERT · FEVER symmetric test v2 | accuracy: +9.05 points | Helped |
| Use CICR-SUM instead of BERT | BERT · FEVER symmetric test v2 | accuracy: +7.77 points | Helped |
| Add causal intervention during training (ablation) | CICR-FC · SNLI hard | accuracy: 83.72 -> 84.64 | Helped |
| Add counterfactual reasoning at inference (ablation) | CICR-FC · SNLI hard | accuracy: 82.80 -> 84.64 | Helped |
| Add causal intervention during training (ablation) | CICR-SUM · SNLI hard | accuracy: 84.33 -> 85.83 | Helped |
| Add counterfactual reasoning at inference (ablation) | CICR-SUM · SNLI hard | accuracy: 83.42 -> 85.83 | Helped |
| Add causal intervention during training (ablation) | CICR-FC · FEVER symmetric test v1 | accuracy: 69.60+13.11 -> 70.01+13.52 | Helped |
| Add counterfactual reasoning at inference (ablation) | CICR-FC · FEVER symmetric test v1 | accuracy: 67.73+12.24 -> 70.01+13.52 | Helped |
Closest papers by meaning
- Debiasing Algorithm through Model Adaptation (ICLR 2024 · similarity 0.88)
- Biases Mitigation and Expressiveness Preservation in Language Models: A Comprehensive Pipeline (Student Abstract) (AAAI 2024 · similarity 0.88)
- Can Large Language Models Infer Causation from Correlation? (ICLR 2024 · similarity 0.87)
- Self-correcting for Debiasing Large Language Models (ICML 2026 · similarity 0.87)
- Prompting Fairness: Integrating Causality to Debias Large Language Models (ICLR 2025 · similarity 0.86)
- LIDAO: Towards Limited Interventions for Debiasing (Large) Language Models (ICML 2024 · similarity 0.86)
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.