AI Paper Atlas

Research map / Large language models / Alignment and preferences

Transitivity Meets Cyclicity: Explicit Preference Decomposition for Dynamic Large Language Model Alignment

Yucong Huang, Xiucheng Li, Kaiqi Zhao, Jing Li

ICML 2026 · Large language models › Alignment and preferences · approach: Robust and calibrated DPO · 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
replace BT/GPM with HRC (dim 2+1) as preference modelreward model · RewardBench 2 averageaccuracy: 56.40 -> 57.63Improved
replace BT/GPM with HRC (dim 2+1) as preference modelLlama-3.1-8B-Instruct reward model · RewardBench 2 averageaccuracy: 70.10 -> 70.95Improved
use HRC instead of BT or GPMreward model, Llama-3.1-8B-Instruct · RewardBench 2 Tiesaccuracy: 73.53 -> 74.51Improved
add context-aware gating (ablation)HRC (Gemma-2B-it, dim 2+1) · RewardBench 2 averageaccuracy: 56.49 -> 57.63Improved
add reward clipping (ablation)HRC (Gemma-2B-it, dim 4+1) · RewardBench 2 averageaccuracy: 57.24 -> 57.12Hurt
add unit-norm constraint (ablation)HRC (Llama-3.1-8B-Instruct, dim 4+1) · RewardBench 2 averageaccuracy: 71.25 -> 70.67Hurt
add unit-norm constraint (ablation)HRC (Llama-3.1-8B-Instruct, dim 2+1) · RewardBench 2 averageaccuracy: 69.98 -> 70.95Improved
add context-aware gating (ablation)HRC (Llama-3.1-8B-Instruct, dim 4+1) · RewardBench 2 averageaccuracy: 71.04 -> 70.67Hurt
switch from static SPPO to DSPPO with HRC oracleGemma-2B-it alignment · AlpacaEval 2.0LC win rate (%): 43.00 -> 44.75Improved
switch from static SPPO to DSPPO with HRC oracleLlama-3.1-8B-Instruct alignment, iteration 2 · MT-Benchscore: 8.42 -> 8.14Hurt
use HRC+DSPPO instead of BT+SPPO or GPM+SPPOGemma-2B-it alignment · Arena-Hard-v0.1win rate (%): 42.1 -> 46.8Improved
transfer HRC+DSPPO preference models to Gemma-2-9B-itGemma-2-9B-it alignment · AlpacaEval 2.0LC win rate (%): 48.79 -> 52.20Improved

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