AI Paper Atlas

Research map / Large language models / Alignment and preferences

Beyond Pairwise: Empowering LLM Alignment With (Ranked) Choice Modeling

Yuxuan Tang, Yifan Feng

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
train on top-2 ranked feedback instead of single-best (ablation)Mallows-RMJ-PO · AlpacaEval 2win rate: 51.17 -> 53.01Improved
use Mallows-RMJ instead of MNL on pairwise data (ablation)RCPO · AlpacaEval 2win rate: 48.08 -> 48.71 (MNL-PO-Discrete vs Mallows-RMJ-PO-Pairwise)Improved
use Mallows-RMJ instead of MNL on top-2 feedback (ablation)RCPO · AlpacaEval 2win rate: 47.69 -> 53.01Improved
raise ranked choice length k beyond 2 (assortment size 5) (ablation)Mallows-RMJ-PO · AlpacaEval 2LC win rate: 41.95 -> 39.92 (k=2 to k=4)Hurt
increase assortment size from 2 to 3 (ablation)Mallows-RMJ-PO · AlpacaEval 2win rate: 48.71 -> 50.49Improved
increase assortment size from 3 to 5 (ablation)Mallows-RMJ-PO-Top-2 · AlpacaEval 2win rate: 49.65 -> 53.01Improved
increase assortment size from 2 to 5 (ablation)Mallows-RMJ-PO · Training timehours: 1.5 -> 3.5Hurt
increase k (ablation)Mallows-RMJ-PO · Training timehoursNo clear effect
train on ranked choice data instead of converting top-2 ranking to all pairs for DPODPO · AlpacaEval 2LC win rate: 33.02 -> 41.95 (DPO-AllPairs vs Mallows-RMJ-PO-Top-2)Improved
apply Mallows-RMJ-PO-Top-2 to Gemma-2-9B-itGemma-2-9B-it · AlpacaEval 2LC win rate: 58.01 (DPO) vs 55.64Mixed
apply Mallows-RMJ-PO-Top-2 to Mistral-7B-InstructMistral-7B-Instruct · Arena-Hardwin rate: 16.8 (DPO) vs 16.9No clear effect
Mallows-RMJ-PO-Top-2 vs SimPO on AlpacaEval LCLlama-3-8B-Instruct · AlpacaEval 2LC win rate: 44.15 -> 41.95Hurt

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