AI Paper Atlas

Research map / Large language models / Alignment and preferences

Multi-Metric Preference Alignment for Generative Speech Restoration

Junan Zhang, Xueyao Zhang, Jing Yang, Yuancheng Wang, Fan Fan, Zhizheng Wu

AAAI 2026 · Large language models › Alignment and preferences · approach: Robust and calibrated DPO · 7 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
apply multi-metric DPO on GenSR-PrefAnyEnhance (MGM, 69k pairs) · Librivox-GSRNISQA: 4.346 -> 4.865Improved
apply multi-metric DPO with only ~3k pairsAR+Soundstorm · Librivox-GSRNISQA: 4.535 -> 4.923Improved
apply multi-metric DPO with ~3k pairsFlow-SR · Librivox-GSRNISQA: 4.184 -> 4.825Improved
DPO alignmentAnyEnhance · Librivox-GSRSpeaker similarity: 0.955 -> 0.954No clear effect
DPO alignment generalizes to DNS speech enhancementAnyEnhance · DNS With ReverbOVRL: 3.204 -> 3.438Improved
use unanimous multi-metric preference pairs instead of single-metric pairs (ablation)AR+Soundstorm DPO · Voicefixer-GSRSIG/OVRL: SIG 3.564 vs 3.531 (NISQA-only), 3.537 (SIM-only)Improved
single-metric NISQA preference pairs (ablation)AR+Soundstorm DPO · Voicefixer-GSRSIG: 3.550 -> 3.531Hurt
use DPO instead of SFT on ground truth or winner samples (ablation)AnyEnhance (MGM) · Librivox-GSR training curvesall metricsImproved
use ground-truth audio as fixed DPO winner (ablation)AnyEnhance (MGM) · training curvesmodel stabilityHurt
use in-paradigm preference data instead of cross-paradigm (ablation)AnyEnhance (MGM) · Librivox-GSROVRL: 3.232 (FM data) -> 3.409 (own data)Improved
fine-tune Voicefixer on DPO-aligned model pseudo-labelsVoicefixer · real-world singing recordingsOVRL: 2.295 -> 2.756Improved

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