PLC-DPO: Posterior Label Correction in Noisy and Ambiguous Preference Optimization

Hugging Face Daily Papers Papers

Summary

PLC-DPO enhances Direct Preference Optimization by routing noisy preference labels into clean, flipped, or tied cases using policy-reference margins, leading to improved performance across various benchmarks.

Direct Preference Optimization (DPO) simplifies alignment through pairwise comparisons but assumes all observed preferences are reliable. Real data often violates this assumption, leading to reversed, weak, or ambiguous labels that cause harmful policy updates. To address this, we propose Posterior Label Correction DPO (PLC-DPO) to robustly optimize preferences by routing each pair's training signal as a clean, flip, or tie case. The key idea is to use the calibrated policy-reference margin as online evidence to take appropriate correction actions. This reframes noisy preference learning as actively correcting supervision direction and strength rather than merely filtering suspicious examples. Across 57 dataset-model-benchmark cells, PLC-DPO obtains the best mean win rate against DPO (60.5 vs. 55.5 for the next-best method). Injected-noise and tie stress tests, human disagreement analysis, and self-confirmation diagnostics further show that the routing remains stable and distinguishes flipped from weakly directional pairs.
Original Article
View Cached Full Text

Cached at: 09/14/26, 02:33 AM

Paper page - PLC-DPO: Posterior Label Correction in Noisy and Ambiguous Preference Optimization

Source: https://huggingface.co/papers/2608.30597

Abstract

Posterior Label Correction DPO improves preference optimization by routing noisy pairwise labels into clean, flipped, or tied cases using calibrated policy-reference margins.

Direct Preference Optimization(DPO) simplifies alignment through pairwise comparisons but assumes all observed preferences are reliable. Real data often violates this assumption, leading to reversed, weak, or ambiguous labels that cause harmful policy updates. To address this, we proposePosterior Label Correction DPO(PLC-DPO) to robustly optimize preferences by routing each pair’s training signal as a clean, flip, or tie case. The key idea is to use the calibratedpolicy-reference marginas online evidence to take appropriate correction actions. This reframes noisypreference learningas actively correcting supervision direction and strength rather than merely filtering suspicious examples. Across 57 dataset-model-benchmark cells, PLC-DPO obtains the best mean win rate against DPO (60.5 vs. 55.5 for the next-best method). Injected-noise and tie stress tests, human disagreement analysis, and self-confirmation diagnostics further show that the routing remains stable and distinguishes flipped from weakly directional pairs.

View arXiv pageView PDFGitHub0Add to collection

Get this paper in your agent:

hf papers read 2608\.30597

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.30597 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.30597 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.30597 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

$\xi$-DPO: Direct Preference Optimization via Ratio Reward Margin

arXiv cs.LG

This paper introduces xi-DPO, a novel preference optimization method that reformulates the objective to minimize distance to optimal ratio reward margins, addressing hyperparameter tuning challenges in SimPO. Experimental results show that xi-DPO outperforms existing methods on open benchmarks.

GroupDPO: Memory efficient Group-wise Direct Preference Optimization

arXiv cs.CL

GroupDPO introduces a memory-efficient algorithm for group-wise direct preference optimization that leverages multiple candidate responses per prompt while reducing peak memory usage through decoupled backpropagation. The method demonstrates consistent improvements over standard DPO across offline and online alignment settings.

Conditional Equivalence of DPO and RLHF: Implicit Assumption, Failure Modes, and Provable Alignment

arXiv cs.AI

This paper proves that the equivalence between Direct Preference Optimization (DPO) and Reinforcement Learning from Human Feedback (RLHF) is conditional and often violated in practice, revealing failure modes where DPO optimizes relative advantage rather than absolute alignment. The authors introduce Constrained Preference Optimization (CPO) for provable alignment and demonstrate state-of-the-art performance.