AI Paper Atlas

Research map / Large language models / Alignment and preferences

$\textit{S}$-SPPO: Semantic-Calibrated Self-Play Preference Optimization

Xiwen Chen, Wenhui Zhu, Jingjing Wang, Peijie Qiu, Zhipeng Wang, Huayu Li, ZhengXiao He, XUANZHAO DONG, Prayag Tiwari, Mingkun Xu, Yujian Xiong, Feng Luo and 4 more

ICML 2026 · Large language models › Alignment and preferences · approach: DPO design choices

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 supervision calibration (semantic gating of win-rate targets) (ablation)SPPO, Llama-3-8B · AlpacaEval 2.0 LC win rate, Iter 3LC win rate: 44.79 -> 46.16Improved
add representation calibration (latent repulsion) (ablation)SPPO, Llama-3-8B · AlpacaEval 2.0 LC win rate, Iter 3LC win rate: 44.79 -> 45.20Improved
combine both calibrations (ablation)SPPO, Llama-3-8B · AlpacaEval 2.0 LC win rate, Iter 3LC win rate: 44.79 -> 47.46Improved
use S-SPPO instead of SPPOSPPO, Llama-3-8B · Arena-Hard-Autowin rate: 29.8 -> 31.5Improved
use S-SPPO instead of SPPOSPPO, Llama-3-8B · MT-Benchaverage score: 7.93 -> 8.22Improved
use S-SPPO instead of SPPOSPPO, Mistral-7B · AlpacaEval 2.0 (GPT-4 annotator)LC win rate: 28.53 -> 30.84Improved
use S-SPPO instead of SPPO in Mistral-7B iteration 1SPPO, Mistral-7B · AlpacaEval 2.0 (GPT-4 annotator)LC win rate: 24.79 -> 24.28Hurt
swap semantic encoder (jina, nomic, MiniLM) (ablation)S-SPPO · AlpacaEval 2.0 LC win rate, Iter 3LC win rate: 46.9 to 47.7No clear effect
vary lambda of representation calibration (ablation)S-SPPO · AlpacaEval 2.0win rateMixed
use S-SPPO instead of SPPOSPPO, Llama-3-8B · GSM8K strict-matchaccuracy: 78.2% -> 78.9%Improved
add S-SPPO calibration overheadSPPO, Llama-3-8B · compute per iterationpeak GPU memory: 67.0GB -> 72.8GBHurt

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