AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

LEC: Linear Expectation Constraints for Selection-Conditioned Risk Control in Selective Prediction and Routing Systems

Zhiyuan Wang, Aniri, Tianlong Chen, Yue Zhang, Heng Tao Shen, Xiaoshuang Shi, Kaidi Xu

ICML 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Algorithms with explicit predictors · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add LEC calibration instead of UCB-CLPsingle-model selective prediction · TriviaQA, Qwen2.5-14B, alpha=0.05power: 0.6240 -> 0.7193Improved
add LEC calibration instead of UCB-HFDsingle-model selective prediction · TriviaQA, Qwen2.5-14B, alpha=0.05power: no feasible threshold -> 0.7193Improved
use LEC instead of UCB-CLPsingle-model selective prediction · TriviaQA, Qwen2.5-3B, alpha=0.1power: 0.5219 -> 0.5998Improved
use LEC instead of UCB-CLPsingle-model selective prediction · TriviaQA, Qwen2.5-7B, alpha=0.05power: 0.3905 -> 0.4889Improved
use LEC instead of UCB-CLPsingle-model selective prediction · TriviaQA, Vicuna-7B-V1.5, alpha=0.1power: 0.5686 -> 0.6508Improved
use LEC instead of UCB-CLP at relaxed risk levelsingle-model selective prediction · TriviaQA, LLaMA-3.1-70B, alpha=0.1power: 1.0 -> 1.0No clear effect
use LEC instead of UCB-CLPsingle-model selective prediction · TriviaQA, Qwen2.5-3Bempirical accepted error rate: 0.0878 -> 0.0987Improved
add routing to LLaMA-3.1-8BLEC with Qwen2.5-3B primary · CommonsenseQA, alpha=0.05accepted sample fraction (%): 20.3% -> 33.9%Improved
add routing to LLaMA-3.1-8BLEC with Qwen2.5-7B primary · CommonsenseQA, alpha=0.05accepted sample fraction (%): 50.7% -> 51.0%No clear effect
add two-model routingLEC with Qwen2.5-3B primary · CommonsenseQA, alpha=0.05accepted correct samples: 965 -> 1610Improved
add two-model routingLEC with Qwen2.5-7B primary · CommonsenseQA, alpha=0.05accepted correct samples: 2392 -> 2413Improved
add joint threshold calibration (ablation)LEC-Routing · CommonsenseQA two-model routingsystem-level error control validityImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy