Research map / Probabilistic and scientific modeling / Molecules and proteins
Multiple-Prediction-Powered Inference
Charlie Cowen-Breen, Alekh Agarwal, Stephen Bates, William W. Cohen, Jacob Eisenstein, Amir Globerson, Adam Fisch
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 |
|---|---|---|---|
| allocate budget over subsets of predictors with MultiPPI instead of best baseline | PPI++ baselines · Chatbot Arena win-rate (Claude-2.1 vs GPT-4-1106) | MSE and CI width | Improved |
| allocate budget with MultiPPI instead of best baseline | PPI++ baselines · ProcessBench (autorater thinking budgets) | MSE and CI width | Improved |
| allocate budget with MultiPPI instead of best baseline | PPI++ baselines · Biography factuality (multi-autorater debate) | MSE and CI width | Improved |
| use vector PPI++ with all models instead of single cheap model | PPI++ · ProcessBench | MSE | Mixed |
| use PPI++ with expensive Gemini 2.5 Pro instead of Flash | PPI++ · Chatbot Arena | MSE | Hurt |
| let the autorater think longer (more words) | autorater · ProcessBench | autorater bias | No clear effect |
| increase number of debate agents vs number of rounds (ablation) | Gemini 2.0 Flash Lite autorater · Biography factuality | autorater accuracy | Improved |
| estimate covariance with Ledoit-Wolf (ablation) | MultiPPI · LLM evaluation experiments | estimation error | Improved |
Closest papers by meaning
- Adaptive Prediction-Powered AutoEval with Reliability and Efficiency Guarantees (NeurIPS 2025 · similarity 0.87)
- Prediction-Powered Inference Across Many Tasks for AI Evaluations and Social Science Research (NeurIPS 2026 · similarity 0.87)
- Prediction-Powered Adaptive Shrinkage Estimation (ICML 2025 · similarity 0.86)
- Scaling Laws Meet Model Architecture: Toward Inference-Efficient LLMs (ICLR 2026 · similarity 0.86)
- Stratified Prediction-Powered Inference for Effective Hybrid Evaluation of Language Models (NeurIPS 2024 · similarity 0.86)
- Cost-of-Pass: An Economic Framework for Evaluating Language Models (ICLR 2026 · 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.
