AI Paper Atlas

Research map / Large language models / Alignment and preferences

Robust Preference Alignment via Directional Neighborhood Consensus

Ruochen Mao, Yuling Shi, Xiaodong Gu, Jiaheng Wei

ICLR 2026 · Large language models › Alignment and preferences · approach: Robust and calibrated DPO

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
sample candidates from neighboring preference directions instead of the target directionDPA-v1-Mistral-7B · UltraFeedback, 45 degRPS win rate (%): 53.4 (20 deg) -> 69.1 (45 deg)Improved
neighborhood sampling (RPS) on SFT modelMistral-7B-Instruct-v0.2 · HelpSteer2, 45 degRPS win rate (%): 52.1 (10 deg) -> 94.3 (45 deg)Improved
neighborhood sampling (RPS) on DPO modelZephyr-7B-Beta · UltraFeedback (mean)RPS win rate (%): 52.1Improved
increase neighborhood size k from 3 to 5 (ablation)RPS on DPA · HelpSteer2 (v3-v10)RPS win rate (0-1): 0.543 -> 0.608Improved
increase neighborhood size k from 3 to 5 (ablation)RPS on SFT · HelpSteer2 (v3-v10)RPS win rate (0-1): 0.564 -> 0.673Improved
increase neighborhood size k from 3 to 5 (ablation)RPS on DPO · HelpSteer2 (v3-v10)RPS win rate (0-1): 0.494 -> 0.535Improved
use angular radius 30 deg instead of 20 deg (ablation)RPS on SFT · HelpSteer2 (v3-v10)RPS win rate (0-1): 0.557 -> 0.673Improved
use angular radius 30 deg instead of 40 deg (ablation)RPS on SFT · HelpSteer2 (v3-v10)RPS win rate (0-1): 0.559 -> 0.673Improved
use angular radius 30 deg instead of 40 deg (ablation)RPS on DPA · HelpSteer2 (v3-v10)RPS win rate (0-1): 0.579 -> 0.608Improved
use angular radius 20 deg instead of 30 deg (ablation)RPS on DPO · HelpSteer2 (v3-v10)RPS win rate (0-1): 0.535 -> 0.485Hurt

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