More Correct Mass, Worse Answers: Why Power Sampling Can Fail and How to Fix It
Power Sampling sharpens a language model's distribution over complete generation trajectories, offering a verifier-free way to improve reasoning at inference time. It also has the potential to serve as a general-purpose front end for a broad range of downstream sampling methods. However, we uncover a striking paradox: Power Sampling can drive more probability mass toward correct trajectories while degrading the downstream inference it is intended to enhance. Using self-consistency as a representative case, we observe accuracy drops of up to 18.5 percentage points across models and reasoning be
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- FuzzyOverlapping authors or contributors · 62%google-research/google-research →
“Shared author/contributor keys: sun”
- LinkedLinked via arxiv author · 85%Haohui Yang →
“More Correct Mass, Worse Answers: Why Power Sampling Can Fail and How to Fix It”
- LinkedLinked via arxiv author · 85%Jiaxing Sun →
“More Correct Mass, Worse Answers: Why Power Sampling Can Fail and How to Fix It”
- LinkedLinked via arxiv author · 85%Xiujun Ma →
“More Correct Mass, Worse Answers: Why Power Sampling Can Fail and How to Fix It”