AI Paper Atlas

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

PETS: A Principled Framework Towards Optimal Trajectory Allocation for Efficient Test-Time Self-Consistency

Zhangyi Liu, Huaizhi Qu, Xiaowei Yin, He Sun, Yanjun Han, Tianlong Chen, Zhun Deng

ICML 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Determinantal point process inference · 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
use PETS-Offline allocation instead of uniformself-consistency sampling · GPQA-Diamond, Qwen3-4Btraces to full self-consistency: 11013 -> 2780Improved
use PETS-Offline allocation instead of uniformself-consistency sampling · AIME25, Qwen3-4Btraces to full self-consistency: 470 -> 212Improved
use PETS-Offline allocation instead of confidence-guidedDeepConf allocation · GPQA-Diamond, Qwen3-4Btraces to full self-consistency: 10652 -> 2780Improved
use PETS-Online allocation instead of uniformself-consistency sampling · GPQA-Diamond, Qwen3-4Btraces to full self-consistency: 9520 -> 4662Improved
use PETS-Online instead of uniformself-consistency sampling · AIME25, gpt-oss-120btraces to full self-consistency: 416 -> 454Hurt
use PETS-Online instead of uniformself-consistency sampling · AIME25, gpt-oss-20btraces to full self-consistency: 221 -> 195Improved
allocate by PETS-Offline instead of uniform at matched tracesself-consistency sampling · GPQA-Diamond, Qwen3-4Baccuracy: 68.9 -> 69.7Improved
add confidence-weighted voting to PETS-OfflinePETS-Offline · GPQA-Diamond, Qwen3-4Btraces to full self-consistency: 2780 -> 3667Hurt
add confidence-weighted voting to PETS-OfflinePETS-Offline · GPQA-Diamond, Qwen3-4Baccuracy: 69.7 -> 70.3Improved
vary number of difficulty bins k from 2 to 5 (ablation)PETS-Online predictor · GPQA-Diamond, Qwen3-4Bpredictor accuracy: 0.6845 -> 0.6786No clear effect
vary KMeans cluster number k from 2 to 8 (ablation)PETS-Oracle · GPQA-Diamond, gpt-oss-120boracle accuracy: 0.8036 -> 0.8274No clear effect
increase predictor bins from 2 to 3 (ablation)PETS-Online predictor · AIME 2024, Qwen3-4Bpredictor accuracy: 0.9500 -> 0.9000Hurt

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