Enregistré dans:
Détails bibliographiques
Auteurs principaux: Wang, Kaiwen, Kidambi, Rahul, Sullivan, Ryan, Agarwal, Alekh, Dann, Christoph, Michi, Andrea, Gelmi, Marco, Li, Yunxuan, Gupta, Raghav, Dubey, Avinava, Ramé, Alexandre, Ferret, Johan, Cideron, Geoffrey, Hou, Le, Yu, Hongkun, Ahmed, Amr, Mehta, Aranyak, Hussenot, Léonard, Bachem, Olivier, Leurent, Edouard
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2407.15762
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909359502524416
author Wang, Kaiwen
Kidambi, Rahul
Sullivan, Ryan
Agarwal, Alekh
Dann, Christoph
Michi, Andrea
Gelmi, Marco
Li, Yunxuan
Gupta, Raghav
Dubey, Avinava
Ramé, Alexandre
Ferret, Johan
Cideron, Geoffrey
Hou, Le
Yu, Hongkun
Ahmed, Amr
Mehta, Aranyak
Hussenot, Léonard
Bachem, Olivier
Leurent, Edouard
author_facet Wang, Kaiwen
Kidambi, Rahul
Sullivan, Ryan
Agarwal, Alekh
Dann, Christoph
Michi, Andrea
Gelmi, Marco
Li, Yunxuan
Gupta, Raghav
Dubey, Avinava
Ramé, Alexandre
Ferret, Johan
Cideron, Geoffrey
Hou, Le
Yu, Hongkun
Ahmed, Amr
Mehta, Aranyak
Hussenot, Léonard
Bachem, Olivier
Leurent, Edouard
contents Reward-based finetuning is crucial for aligning language policies with intended behaviors (e.g., creativity and safety). A key challenge is to develop steerable language models that trade-off multiple (conflicting) objectives in a flexible and efficient manner. This paper presents Conditional Language Policy (CLP), a general framework for finetuning language models on multiple objectives. Building on techniques from multi-task training and parameter-efficient finetuning, CLP learn steerable models that effectively trade-off conflicting objectives at inference time. Notably, this does not require training or maintaining multiple models to achieve different trade-offs between the objectives. Through extensive experiments and ablations on two summarization datasets, we show that CLP learns steerable language models that outperform and Pareto-dominate the existing approaches for multi-objective finetuning.
format Preprint
id arxiv_https___arxiv_org_abs_2407_15762
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Conditional Language Policy: A General Framework for Steerable Multi-Objective Finetuning
Wang, Kaiwen
Kidambi, Rahul
Sullivan, Ryan
Agarwal, Alekh
Dann, Christoph
Michi, Andrea
Gelmi, Marco
Li, Yunxuan
Gupta, Raghav
Dubey, Avinava
Ramé, Alexandre
Ferret, Johan
Cideron, Geoffrey
Hou, Le
Yu, Hongkun
Ahmed, Amr
Mehta, Aranyak
Hussenot, Léonard
Bachem, Olivier
Leurent, Edouard
Machine Learning
Artificial Intelligence
Computation and Language
Reward-based finetuning is crucial for aligning language policies with intended behaviors (e.g., creativity and safety). A key challenge is to develop steerable language models that trade-off multiple (conflicting) objectives in a flexible and efficient manner. This paper presents Conditional Language Policy (CLP), a general framework for finetuning language models on multiple objectives. Building on techniques from multi-task training and parameter-efficient finetuning, CLP learn steerable models that effectively trade-off conflicting objectives at inference time. Notably, this does not require training or maintaining multiple models to achieve different trade-offs between the objectives. Through extensive experiments and ablations on two summarization datasets, we show that CLP learns steerable language models that outperform and Pareto-dominate the existing approaches for multi-objective finetuning.
title Conditional Language Policy: A General Framework for Steerable Multi-Objective Finetuning
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2407.15762