Research map / Label-efficient and robust learning / Recognition and long-tail
Partial Order in Chaos: Consensus on Feature Attributions in the Rashomon Set
Gabriel Laberge, Yann Pequignot, Alexandre Mathieu, Foutse Khomh, Mario Marchand
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 |
|---|---|---|---|
| explain the average model instead of individual models, using aggregated attribution | SHAP on MLP ensemble · Toy regression | feature ordering x1 vs x2: x2 4.59 vs x1 3.98 | Hurt |
| average models into h_E | MLP ensemble · Toy regression | test RMSE: 0.47 -> 0.43 | Improved |
| use ordinal consensus as uncertainty score | mean-rank aggregation · Toy regression | uncertainty calibration: consensus 0.83 for x1 and x2 despite a second-best model disagreeing (RMSE 0.512) | Hurt |
| use variance as uncertainty score | attribution aggregation · Toy regression | uncertainty calibration: variance 0.50 and 0.35 for x2 and x1 though both exceed x3 and x4 for all models | Hurt |
| abstain on feature pairs where good models disagree (partial order of consensus) | Rashomon Set explanations · Toy regression | trustworthy statements | Improved |
| group correlated features (Garage, Bsmt, AbvGrd) and use joint attribution (ablation) | Additive Regression Rashomon Set · Kaggle-Houses | partial order cardinality | Improved |
| add grouping of correlated features to local attributions (ablation) | Additive Regression Rashomon Set · Kaggle-Houses | sign consistency of Garage and Bsmt attributions | Improved |
| vary tolerance epsilon around 0.1444 (ablation) | Additive Regression Rashomon Set · Kaggle-Houses | partial order cardinality | No clear effect |
| compare consensus ranking with min-max intervals of global importance (Model Class Reliance) | global feature importance · Kaggle-Houses | ordering of 1stFlrSF vs OverallCond | Improved |
| increase quadrature steps by factor 10 (ablation) | Integrated Gradient trapezoid · COMPAS | Gap Error: error reduced by a factor 100 | Improved |
| include slightly worse models in the Rashomon Set (epsilon 4.23) (ablation) | Kernel Ridge IG attributions · COMPAS | attribution of Race=Black | Mixed |
| switch polynomial kernel to Gaussian kernel (ablation) | Kernel Ridge Regression · COMPAS | test RMSE: 2.12 -> 2.11 | No clear effect |
Closest papers by meaning
- Regional Explanations: Bridging Local and Global Variable Importance (NeurIPS 2025 · similarity 0.86)
- OrdShap: Feature Position Importance for Sequential Black-Box Models (NeurIPS 2025 · similarity 0.85)
- Exploring the cloud of feature interaction scores in a Rashomon set (ICLR 2024 · similarity 0.85)
- Provably Better Explanations with Optimized Aggregation of Feature Attributions (ICML 2024 · similarity 0.85)
- Local vs. Global Interpretability: A Computational Complexity Perspective (ICML 2024 · similarity 0.84)
- Feature Attribution with Necessity and Sufficiency via Dual-stage Perturbation Test for Causal Explanation (ICML 2024 · 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.
