On Predicting the Post-training Potential of Pre-trained LLMs

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Xiaoyuan, Ma, Yubo, Yang, Kexin, Li, Moxin, Bao, Keqin, Wang, Wenie, Feng, Fuli, Liu, Dayiheng
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911674108215296
author Li, Xiaoyuan
Ma, Yubo
Yang, Kexin
Li, Moxin
Bao, Keqin
Wang, Wenie
Feng, Fuli
Liu, Dayiheng
author_facet Li, Xiaoyuan
Ma, Yubo
Yang, Kexin
Li, Moxin
Bao, Keqin
Wang, Wenie
Feng, Fuli
Liu, Dayiheng
contents The performance of Large Language Models (LLMs) on downstream tasks is fundamentally constrained by the capabilities acquired during pre-training. However, traditional benchmarks like MMLU often fail to reflect a base model's plasticity in complex open-ended scenarios, leading to inefficient model selection. We address this by introducing a new task of predicting post-training potential - forecasting a base model's performance before post-training. We propose RuDE (Rubric-based Discriminative Evaluation), a unified framework that bypasses the generation gap of base models by leveraging response discrimination. Guided by our systematic 4C Taxonomy, RuDE constructs controlled contrastive pairs across diverse domains by fine-grained rubric violations. Extensive experiments demonstrate a correlation greater than 90% with post-training performance. Crucially, validation via Reinforcement Learning (RL) confirms that RuDE effectively identifies high-potential smaller models that outperform larger counterparts, offering a compute-efficient mechanism for foundation model development.
format Preprint
id arxiv_https___arxiv_org_abs_2605_11978
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle On Predicting the Post-training Potential of Pre-trained LLMs
Li, Xiaoyuan
Ma, Yubo
Yang, Kexin
Li, Moxin
Bao, Keqin
Wang, Wenie
Feng, Fuli
Liu, Dayiheng
Computation and Language
The performance of Large Language Models (LLMs) on downstream tasks is fundamentally constrained by the capabilities acquired during pre-training. However, traditional benchmarks like MMLU often fail to reflect a base model's plasticity in complex open-ended scenarios, leading to inefficient model selection. We address this by introducing a new task of predicting post-training potential - forecasting a base model's performance before post-training. We propose RuDE (Rubric-based Discriminative Evaluation), a unified framework that bypasses the generation gap of base models by leveraging response discrimination. Guided by our systematic 4C Taxonomy, RuDE constructs controlled contrastive pairs across diverse domains by fine-grained rubric violations. Extensive experiments demonstrate a correlation greater than 90% with post-training performance. Crucially, validation via Reinforcement Learning (RL) confirms that RuDE effectively identifies high-potential smaller models that outperform larger counterparts, offering a compute-efficient mechanism for foundation model development.
title On Predicting the Post-training Potential of Pre-trained LLMs
topic Computation and Language
url https://arxiv.org/abs/2605.11978