AI Paper Atlas

Research map / Large language models / Alignment and preferences

RSPO: Regularized Self-Play Alignment of Large Language Models

Xiaohang Tang, Sangwoong Yoon, Seongho Son, Huizhuo Yuan, Quanquan Gu, Ilija Bogunovic

ICML 2026 · Large language models › Alignment and preferences · approach: DPO design choices · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add reverse KL regularization to SPPO loss (ablation)SPPO · AlpacaEval 2.0, Mistral-7B, iterationsLC win rate: 28.53 -> 34.21 (Iter3)Improved
add IS-forward plus reverse KL regularization (ablation)SPPO · AlpacaEval 2.0, Mistral-7BLC win rate: 28.53 -> 35.44Improved
add chi-square regularization (ablation)SPPO · AlpacaEval 2.0, Mistral-7B, Iter3LC win rate: 28.53 -> 29.97Improved
increase forward KL regularization temperature (ablation)RSPO (IS-Forward KL) · Mistral-7Bresponse length and win rateHurt
increase reverse KL regularization temperature (ablation)RSPO (Reverse KL) · Mistral-7Bresponse lengthImproved
add reverse KL regularization (ablation)SPPO · Self-BLEU diversity, Mistral-7B Iter3Self-BLEU: 0.758 -> 0.691Improved
run SPPO for more iterations (up to 9) (ablation)SPPO · AlpacaEval 2.0, Mistral-7BLC win rate: 28.53 -> 29.17No clear effect
use stronger preference model GPM-8B instead of PairRM (ablation)SPPO / RSPO on LLaMA-8B · AlpacaEval 2.0LC win rate: +0.44Improved
add RSPO regularization on LLaMA-8BSPPO · AlpacaEval 2.0LC win rate: 38.77 -> 43.66Improved
add RSPO regularization on Gemma-2BSPPO · AlpacaEval 2.0LC win rate: 50.54 -> 51.83Improved

Later papers that compared with this method

1 later accepted paper reports RSPO as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy