Enregistré dans:
Détails bibliographiques
Auteurs principaux: Cao, Sheng, Wu, Mingrui, Prasad, Karthik, Tian, Yuandong, Liu, Zechun
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2504.21023
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866917306776420352
author Cao, Sheng
Wu, Mingrui
Prasad, Karthik
Tian, Yuandong
Liu, Zechun
author_facet Cao, Sheng
Wu, Mingrui
Prasad, Karthik
Tian, Yuandong
Liu, Zechun
contents The post-training phase of large language models is essential for enhancing capabilities such as instruction-following, reasoning, and alignment with human preferences. However, it demands extensive high-quality data and poses risks like overfitting, alongside significant computational costs due to repeated post-training and evaluation after each base model update. This paper introduces $ParamΔ$, a novel method that streamlines post-training by transferring knowledge from an existing post-trained model to a newly updated base model with ZERO additional training. By computing the difference between post-trained model weights ($Θ_\text{post}$) and base model weights ($Θ_\text{base}$), and adding this to the updated base model ($Θ'_\text{base}$), we define $ParamΔ$ Model as: $Θ_{\text{Param}Δ} = Θ_\text{post} - Θ_\text{base} + Θ'_\text{base}$. This approach surprisingly equips the new base model with post-trained capabilities, achieving performance comparable to direct post-training. We did analysis on LLama3, Llama3.1, Qwen, and DeepSeek-distilled models. Results indicate $ParamΔ$ Model effectively replicates traditional post-training. For example, the $ParamΔ$ Model obtained from 70B Llama3-inst, Llama3-base, Llama3.1-base models attains approximately 95\% of Llama3.1-inst model's performance on average. $ParamΔ$ brings a new perspective on how to fully leverage models in the open-weight community, where checkpoints for base and instruct models are readily available and frequently updated, by providing a cost-free framework to accelerate the iterative cycle of model development.
format Preprint
id arxiv_https___arxiv_org_abs_2504_21023
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Param$Δ$ for Direct Weight Mixing: Post-Train Large Language Model at Zero Cost
Cao, Sheng
Wu, Mingrui
Prasad, Karthik
Tian, Yuandong
Liu, Zechun
Computation and Language
Artificial Intelligence
Machine Learning
The post-training phase of large language models is essential for enhancing capabilities such as instruction-following, reasoning, and alignment with human preferences. However, it demands extensive high-quality data and poses risks like overfitting, alongside significant computational costs due to repeated post-training and evaluation after each base model update. This paper introduces $ParamΔ$, a novel method that streamlines post-training by transferring knowledge from an existing post-trained model to a newly updated base model with ZERO additional training. By computing the difference between post-trained model weights ($Θ_\text{post}$) and base model weights ($Θ_\text{base}$), and adding this to the updated base model ($Θ'_\text{base}$), we define $ParamΔ$ Model as: $Θ_{\text{Param}Δ} = Θ_\text{post} - Θ_\text{base} + Θ'_\text{base}$. This approach surprisingly equips the new base model with post-trained capabilities, achieving performance comparable to direct post-training. We did analysis on LLama3, Llama3.1, Qwen, and DeepSeek-distilled models. Results indicate $ParamΔ$ Model effectively replicates traditional post-training. For example, the $ParamΔ$ Model obtained from 70B Llama3-inst, Llama3-base, Llama3.1-base models attains approximately 95\% of Llama3.1-inst model's performance on average. $ParamΔ$ brings a new perspective on how to fully leverage models in the open-weight community, where checkpoints for base and instruct models are readily available and frequently updated, by providing a cost-free framework to accelerate the iterative cycle of model development.
title Param$Δ$ for Direct Weight Mixing: Post-Train Large Language Model at Zero Cost
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2504.21023