AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

DARC: Disagreement-Aware Alignment via Risk-Constrained Decoding

mingxi Zou, Jiaxiang Chen, Junfan Li, Langzhang Liang, Qifan Wang, Xu Yinghui, Zenglin Xu

ICML 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Active learning and subset selection · 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
use DARC-ε instead of mean Best-of-K selectionLlama-3.1-8B-Instruct · MT-Bench human eval (high-disagreement top 20%)human score: 7.62 -> 8.34Improved
add epsilon tie-breaking (least-controversial among near-optimal) to entropic decoding (ablation)DARC · MT-Bench human eval (overall)human score: 7.84 -> 8.08Improved
apply DARC-ε on top of a cDPO policycDPO · MT-Bench human eval (high-disagreement top 20%)disagreement risk: 0.89 -> 0.63Improved
apply DARC-ε on top of rDPO policyrDPO · MT-Bench human eval (overall)human score: 8.17 -> 8.15No clear effect
DeAL lookahead decoding vs Best-of-KLlama-3.1-8B-Instruct · MT-Bench human eval (high-disagreement top 20%)disagreement risk: 1.00 -> 1.13Hurt
increase candidate pool size K (ablation)DARC-ε · MT-BenchΔTradeoff and ΔCVaRImproved
increase risk sensitivity β (ablation)DARC-ε · MT-BenchΔTradeoff and ΔCVaRImproved
enlarge near-optimal set via ε (ablation)DARC-ε · MT-BenchrobustnessMixed
increase perturbation budget Naug (ablation)DARC-ε · MT-Benchrobustness: +1.54% / +1.98% / +3.50% latency for Naug 4/8/16No clear effect
multi-scorer robust aggregation with DARC-εDARC-ε · Multi-scorer, high-σ subsetW/T/L vs Base: 44 / 54 / 2Improved
use DARC-ε instead of Best-of-K on low-disagreement promptsDARC-ε · MT-Bench human bucket Q1human score delta: 0.125; W/T/L 11 / 80 / 9No clear effect
pairwise scalarization instead of absolute ratings (ablation)DARC-ε · MT-Bench human eval, high-disagreementmean ∆: +1.037 (S0) vs +0.321 (S1)No clear effect

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