SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.08005 · ICLR · 2024

Refined Direct Preference Optimization with Synthetic Data for Behavioral Alignment of LLMs

Víctor Gallego

arXiv · PDF · Open in the Atlas

Code that ran

We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

In this paper, we introduce refined Direct Preference Optimization (rDPO), a method for improving the behavioral alignment of Large Language Models (LLMs) without the need for human-annotated data. The method involves creating synthetic data using self-critique prompting by a teacher LLM and then utilising an generalized DPO loss function to distil to a student LLM. The loss function incorporates an additional external reward model to improve the quality of synthetic data, making rDPO robust to potential noise in the synthetic dataset. rDPO is shown to be effective in a diverse set of behavioural alignment tasks, such as improved safety, robustness against role-playing, and reduced sycophancy. Code to be released at github.com/vicgalle/refined-dpo.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2402.08005")
get_code_for_paper("2402.08005")
have("2402.08005")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2402.08005.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2402.08005.svg)](https://syntology.ai/paper/2402.08005)

Connect an agent — have() is free.