Don't Throw Away Your Pretrained Model

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Feng, Shangbin, Yu, Wenhao, Wang, Yike, Zhang, Hongming, Tsvetkov, Yulia, Yu, Dong
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917004730957824
author Feng, Shangbin
Yu, Wenhao
Wang, Yike
Zhang, Hongming
Tsvetkov, Yulia
Yu, Dong
author_facet Feng, Shangbin
Yu, Wenhao
Wang, Yike
Zhang, Hongming
Tsvetkov, Yulia
Yu, Dong
contents Alignment training has tradeoffs: it helps language models (LMs) gain in reasoning and instruction following but might lose out on skills such as creativity and calibration, where unaligned base models are better at. We aim to make the best of both worlds through model collaboration, where different models in the training pipeline collaborate and complement each other. Since LM responses feature interleaving skills that favor different models, we propose Switch Generation, where pretrained and aligned model versions take turns to ``speak'' in a response sequence. Specifically, we train a switcher LM by learning from outcomes of choosing different models to generate the next segment across diverse queries and contexts. At inference time, the switcher LM guides different model checkpoints to dynamically generate the next segment where their strengths are most needed. Extensive experiments with 8 model collaboration baselines and 18 datasets show that 1) model collaboration consistently outperforms individual models on 16 out of 18 tasks, and 2) Switch Generation further outperforms baselines by 12.9% on average. Further analysis reveals that Switch Generation discovers compositional skills to solve problems where individual models struggle and generalizes to unseen models and tasks, reusing and repurposing by-products in expensive model training pipelines that are otherwise discarded.
format Preprint
id arxiv_https___arxiv_org_abs_2510_09913
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Don't Throw Away Your Pretrained Model
Feng, Shangbin
Yu, Wenhao
Wang, Yike
Zhang, Hongming
Tsvetkov, Yulia
Yu, Dong
Computation and Language
Alignment training has tradeoffs: it helps language models (LMs) gain in reasoning and instruction following but might lose out on skills such as creativity and calibration, where unaligned base models are better at. We aim to make the best of both worlds through model collaboration, where different models in the training pipeline collaborate and complement each other. Since LM responses feature interleaving skills that favor different models, we propose Switch Generation, where pretrained and aligned model versions take turns to ``speak'' in a response sequence. Specifically, we train a switcher LM by learning from outcomes of choosing different models to generate the next segment across diverse queries and contexts. At inference time, the switcher LM guides different model checkpoints to dynamically generate the next segment where their strengths are most needed. Extensive experiments with 8 model collaboration baselines and 18 datasets show that 1) model collaboration consistently outperforms individual models on 16 out of 18 tasks, and 2) Switch Generation further outperforms baselines by 12.9% on average. Further analysis reveals that Switch Generation discovers compositional skills to solve problems where individual models struggle and generalizes to unseen models and tasks, reusing and repurposing by-products in expensive model training pipelines that are otherwise discarded.
title Don't Throw Away Your Pretrained Model
topic Computation and Language
url https://arxiv.org/abs/2510.09913