Research map / Learning theory and algorithms / Prediction and decision losses
FOCUS: Flexible Optimizable Counterfactual Explanations for Tree Ensembles
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 FOCUS instead of DACE for Mahalanobis counterfactual generation | Tree-based classifiers · Wine Quality, HELOC, COMPAS, Shopping (six runnable settings) | mean distance: 1.325 -> 0.542; 1.427 -> 0.810; 0.814 -> 0.776; 1.570 -> 0.636; 0.050 -> 0.023; 3.230 -> 0.303 | Helped |
| Use FOCUS instead of Feature Tweaking | Tree ensembles across 36 dataset/model/distance settings · Wine Quality, HELOC, COMPAS, Shopping | mean distance: FOCUS outperforms FT in 20 settings; FT outperforms FOCUS in 8; no significant difference in 8 | Mixed |
| Use FOCUS instead of Feature Tweaking with Cosine distance | Tree-based classifiers · Across the four datasets | mean distance | Helped |
| Use FOCUS instead of Feature Tweaking with Manhattan distance | Tree-based classifiers · Across the four datasets | mean distance | Mixed |
| Use FOCUS instead of Feature Tweaking | Random Forest and Adaptive Boosting models · Across tested settings | mean distance | Helped |
| Use FOCUS instead of Feature Tweaking | Decision Tree models · Across tested settings | mean distance | Hurt |
| Use FOCUS rather than Feature Tweaking | Tree-based classifiers · All experimental settings | counterfactual coverage: every instance | Helped |
| Use FOCUS rather than DACE | Tree-based classifiers · Test instances and models that exceed DACE's supported problem size | counterfactual coverage: 1,470; 3,138; 1,852; 1,852; 3,699; 3,699 | Helped |
| Replace hard tree split indicators with sigmoid split approximations and ensemble maximum with softmax | FOCUS counterfactual optimization · Tree ensembles | mean distance | Helped |
Closest papers by meaning
- Counterfactual Maps: What They Are and How to Find Them (NeurIPS 2026 · similarity 0.90)
- Generating Likely Counterfactuals Using Sum-Product Networks (ICLR 2025 · similarity 0.88)
- A General Search-Based Framework for Generating Textual Counterfactual Explanations (AAAI 2024 · similarity 0.87)
- Promoting Counterfactual Robustness through Diversity (AAAI 2024 · similarity 0.86)
- What makes an Ensemble (Un) Interpretable? (ICML 2025 · similarity 0.86)
- Faithful Model Explanations through Energy-Constrained Conformal Counterfactuals (AAAI 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.