Latent Preference Coding: Aligning Large Language Models via Discrete Latent Codes

Zhuocheng Gong, Jian Guan, Wei Wu, Huishuai Zhang, Dongyan Zhao
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:19897-19911, 2025.

Abstract

Large language models (LLMs) have achieved remarkable success, yet aligning their generations with human preferences remains a critical challenge. Existing approaches to preference modeling often rely on an explicit or implicit reward function, overlooking the intricate and multifaceted nature of human preferences that may encompass conflicting factors across diverse tasks and populations. To address this limitation, we introduce Latent Preference Coding (LPC), a novel framework that models the implicit factors as well as their combinations behind holistic preferences using discrete latent codes. LPC seamlessly integrates with various offline alignment algorithms, automatically inferring the underlying factors and their importance from data without relying on pre-defined reward functions and hand-crafted combination weights. Extensive experiments on multiple benchmarks demonstrate that LPC consistently improves upon three alignment algorithms (DPO, SimPO, and IPO) using three base models (Mistral-7B, Llama3-8B, and Llama3-Instruct-8B). Furthermore, deeper analysis reveals that the learned latent codes effectively capture the differences in the distribution of human preferences and significantly enhance the robustness of alignment algorithms against noise in data. By providing a unified representation for the multifarious preference factors, LPC paves the way towards developing more robust and versatile alignment techniques for responsible deployment of powerful LLMs.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-gong25a, title = {Latent Preference Coding: Aligning Large Language Models via Discrete Latent Codes}, author = {Gong, Zhuocheng and Guan, Jian and Wu, Wei and Zhang, Huishuai and Zhao, Dongyan}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {19897--19911}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/gong25a/gong25a.pdf}, url = {https://proceedings.mlr.press/v267/gong25a.html}, abstract = {Large language models (LLMs) have achieved remarkable success, yet aligning their generations with human preferences remains a critical challenge. Existing approaches to preference modeling often rely on an explicit or implicit reward function, overlooking the intricate and multifaceted nature of human preferences that may encompass conflicting factors across diverse tasks and populations. To address this limitation, we introduce Latent Preference Coding (LPC), a novel framework that models the implicit factors as well as their combinations behind holistic preferences using discrete latent codes. LPC seamlessly integrates with various offline alignment algorithms, automatically inferring the underlying factors and their importance from data without relying on pre-defined reward functions and hand-crafted combination weights. Extensive experiments on multiple benchmarks demonstrate that LPC consistently improves upon three alignment algorithms (DPO, SimPO, and IPO) using three base models (Mistral-7B, Llama3-8B, and Llama3-Instruct-8B). Furthermore, deeper analysis reveals that the learned latent codes effectively capture the differences in the distribution of human preferences and significantly enhance the robustness of alignment algorithms against noise in data. By providing a unified representation for the multifarious preference factors, LPC paves the way towards developing more robust and versatile alignment techniques for responsible deployment of powerful LLMs.} }
Endnote
%0 Conference Paper %T Latent Preference Coding: Aligning Large Language Models via Discrete Latent Codes %A Zhuocheng Gong %A Jian Guan %A Wei Wu %A Huishuai Zhang %A Dongyan Zhao %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-gong25a %I PMLR %P 19897--19911 %U https://proceedings.mlr.press/v267/gong25a.html %V 267 %X Large language models (LLMs) have achieved remarkable success, yet aligning their generations with human preferences remains a critical challenge. Existing approaches to preference modeling often rely on an explicit or implicit reward function, overlooking the intricate and multifaceted nature of human preferences that may encompass conflicting factors across diverse tasks and populations. To address this limitation, we introduce Latent Preference Coding (LPC), a novel framework that models the implicit factors as well as their combinations behind holistic preferences using discrete latent codes. LPC seamlessly integrates with various offline alignment algorithms, automatically inferring the underlying factors and their importance from data without relying on pre-defined reward functions and hand-crafted combination weights. Extensive experiments on multiple benchmarks demonstrate that LPC consistently improves upon three alignment algorithms (DPO, SimPO, and IPO) using three base models (Mistral-7B, Llama3-8B, and Llama3-Instruct-8B). Furthermore, deeper analysis reveals that the learned latent codes effectively capture the differences in the distribution of human preferences and significantly enhance the robustness of alignment algorithms against noise in data. By providing a unified representation for the multifarious preference factors, LPC paves the way towards developing more robust and versatile alignment techniques for responsible deployment of powerful LLMs.
APA
Gong, Z., Guan, J., Wu, W., Zhang, H. & Zhao, D.. (2025). Latent Preference Coding: Aligning Large Language Models via Discrete Latent Codes. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:19897-19911 Available from https://proceedings.mlr.press/v267/gong25a.html.

Related Material