Random Is Hard to Beat: Active Selection in online DPO with Modern LLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Oh, Giyeong, Lee, Junghyun, Park, Jaehyun, Yu, Youngjae, Bae, Wonho, Noh, Junhyug
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913001285615616
author Oh, Giyeong
Lee, Junghyun
Park, Jaehyun
Yu, Youngjae
Bae, Wonho
Noh, Junhyug
author_facet Oh, Giyeong
Lee, Junghyun
Park, Jaehyun
Yu, Youngjae
Bae, Wonho
Noh, Junhyug
contents Modern LLMs inherit strong priors from web-scale pretraining, which can limit the headroom of post-training data-selection strategies. While Active Preference Learning (APL) seeks to optimize query efficiency in online Direct Preference Optimization (DPO), the inherent richness of on-policy candidate pools often renders simple Random sampling a surprisingly formidable baseline. We evaluate uncertainty-based APL against Random across harmlessness, helpfulness, and instruction-following settings, utilizing both reward models and LLM-as-a-judge proxies. We find that APL yields negligible improvements in proxy win-rates compared to Random. Crucially, we observe a dissociation where win-rate improves even as general capability -- measured by standard benchmarks -- degrades. APL fails to mitigate this capability collapse or reduce variance significantly better than random sampling. Our findings suggest that in the regime of strong pre-trained priors, the computational overhead of active selection is difficult to justify against the ``cheap diversity'' provided by simple random samples. Our code is available at https://github.com/BootsofLagrangian/random-vs-apl.
format Preprint
id arxiv_https___arxiv_org_abs_2604_02766
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Random Is Hard to Beat: Active Selection in online DPO with Modern LLMs
Oh, Giyeong
Lee, Junghyun
Park, Jaehyun
Yu, Youngjae
Bae, Wonho
Noh, Junhyug
Machine Learning
Artificial Intelligence
Modern LLMs inherit strong priors from web-scale pretraining, which can limit the headroom of post-training data-selection strategies. While Active Preference Learning (APL) seeks to optimize query efficiency in online Direct Preference Optimization (DPO), the inherent richness of on-policy candidate pools often renders simple Random sampling a surprisingly formidable baseline. We evaluate uncertainty-based APL against Random across harmlessness, helpfulness, and instruction-following settings, utilizing both reward models and LLM-as-a-judge proxies. We find that APL yields negligible improvements in proxy win-rates compared to Random. Crucially, we observe a dissociation where win-rate improves even as general capability -- measured by standard benchmarks -- degrades. APL fails to mitigate this capability collapse or reduce variance significantly better than random sampling. Our findings suggest that in the regime of strong pre-trained priors, the computational overhead of active selection is difficult to justify against the ``cheap diversity'' provided by simple random samples. Our code is available at https://github.com/BootsofLagrangian/random-vs-apl.
title Random Is Hard to Beat: Active Selection in online DPO with Modern LLMs
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2604.02766