Score-Repellent Monte Carlo: Toward Efficient Non-Markovian Sampler with Constant Memory in General State Spaces

Jie Hu, Lingyun Chen, Geeho Kim, Jinyoung Choi, Bohyung Han, Do Young Eun · 2026

Read Score-Repellent Monte Carlo: Toward Efficient Non-Markovian Sampler with Constant Memory in General State Spaces by Jie Hu, Lingyun Chen, Geeho Kim, Jinyoung Choi, Bohyung Han, Do Young Eun (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-22948.txt.

Description

Hu and five colleagues propose Score-Repellent Monte Carlo, a history-dependent sampler summing up trajectory history as a running average of score evaluations, not an empirical measure over finite states. That history becomes a surrogate target through an exponential score tilt, letting any base kernel run on it as history updates online. The analysis proves almost sure convergence and a joint central limit theorem, with asymptotic covariance falling as repulsion strength rises. Tests show improved estimator variance and mode coverage, with memory scaling with dimension and modest overhead.

Catalog record

CREATORJie Hu, Lingyun Chen, Geeho Kim, Jinyoung Choi, Bohyung Han, Do Young Eun
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE173263
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.22948

Browse the full library · About the Tower of Babel