Research map / Learning theory and algorithms / Prediction and decision losses
Consensual Affine Transformations for Partial Valuation Aggregation
Hermann Schichl, Meinolf Sellmann
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 |
|---|---|---|---|
| drop range-matching term (NoRange vs Aff-NLP) (ablation) | consensual affine transformation · 50 obj/30 exp latent metric | error rate (%): 11.4 vs 10.6 | Mixed |
| use iterated LP for range instead of NoRange (ablation) | consensual affine · one true value 50x30 | error rate (%): 0.55 -> 1.27 | Hurt |
| use integer program for range term (Aff-IP) vs NoRange (ablation) | consensual affine · one true value 50x30 | error rate (%): 0.55 -> 0.70 | Hurt |
| replace Borda/average with Aff-NLP | aggregation · 50 obj/30 exp latent metric | error rate (%): 13.7 -> 10.6 | Improved |
| replace Kemeny-Young with Aff-NLP | aggregation · 50 obj/30 exp latent metric | error rate (%): 22.0 -> 10.6 | Improved |
| replace MLE with NoRange on latent-metric data | aggregation · 40 obj/20 exp latent metric | error rate (%): 12.2 -> 9.72 | Improved |
| replace NoRange with MLE on one-true-value data | aggregation · 50 obj/30 exp one true value | error rate (%): 0.23 vs 0.55 | Hurt |
| replace Borda with average on one-true-value data | aggregation · 50 obj/30 exp one true value | error rate (%): 8.46 -> 9.62 | Hurt |
| weighted vs unweighted Kemeny-Young (ablation) | Kemeny-Young · all generated sets | error rate (%): 7.44 vs 7.44 | No clear effect |
| Aff-IP vs Aff-NLP runtime (ablation) | consensual affine · 50 obj/30 exp one true value | CPU time (s): 0.25 -> 47.1 | Hurt |
| Kemeny-Young vs Aff-NLP runtime | aggregation · 50 obj/30 exp latent metric | CPU time (s): 0.14 -> 135K | Hurt |
| Aff-NLP at scale | Aff-NLP · 2000 objects, 500 experts | error rate (%): under 0.6% | Improved |
Closest papers by meaning
- CARE: Confounder-Aware Aggregation for Reliable LLM Evaluation (ICML 2026 · similarity 0.84)
- Limits and Potential of Score-Based Data Valuation: Redundancy, Complementarity, and Non-Monotonicity (NeurIPS 2026 · similarity 0.84)
- Welfare, Improvability, and Variance: A Principal-Agent Approach to Optimal Benchmark Item Aggregation (NeurIPS 2026 · similarity 0.84)
- Selective Preference Aggregation (ICML 2025 · similarity 0.82)
- Aggregating Quantitative Relative Judgments: From Social Choice to Ranking Prediction (NeurIPS 2024 · similarity 0.82)
- DU-Shapley: A Shapley Value Proxy for Efficient Dataset Valuation (NeurIPS 2024 · similarity 0.81)
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.
