paper-with-me

홈 › Papers

A Statistical Framework for Alignment with Biased AI Feedback

2026-02-09 · Xintao Xia, Zhiqiu Xia, Linjun Zhang, Zhanrui Cai arxiv

Modern alignment pipelines are increasingly replacing expensive human preference labels with evaluations from large language models (LLM-as-Judge). However, AI labels can be systematically biased compared to high-quality human feedback datasets. In this paper, we develop two debiased alignment methods within a general framework that accommodates heterogeneous prompt-response distributions and external human feedback sources. Debiased Direct Preference Optimization (DDPO) augments standard DPO with a residual-based correction and density-ratio reweighting to mitigate systematic bias, while retaining DPO's computational efficiency. Debiased Identity Preference Optimization (DIPO) directly estimates human preference probabilities without imposing a parametric reward model. We provide theoretical guarantees for both methods: DDPO offers a practical and computationally efficient solution for large-scale alignment, whereas DIPO serves as a robust, statistically optimal alternative that attains the semiparametric efficiency bound. Empirical studies on sentiment generation, summarization, and single-turn dialogue demonstrate that the proposed methods substantially improve alignment efficiency and recover performance close to that of an oracle trained on fully human-labeled data.

📄 PDF Abstract BibTeX arXiv:2602.08259

Code (0)

등록된 구현이 없습니다.

Tasks

Computational Efficiency

Similar Papers 제목 키워드 기반

Accelerating Unbiased LLM Evaluation via Synthetic Feedback

2025-02-14 · Zhaoyi Zhou, Yuda Song, Andrea Zanette

When developing new large language models (LLMs), a key step is evaluating their final performance, often by computing the win-rate against a reference model based on external feedback. Human feedback is the gold standar…

Unbiased Alignment for Large Language Models with Noisy Preferences

2026-07-03 · Jialiang Wang, Xianming Liu, Xiong Zhou, Hui Liu 외 arxiv

The alignment of large language models with human preferences is commonly achieved through Reinforcement Learning from Human Feedback or Direct Preference Optimization. However, these methods are vulnerable to the signif…

Reinforcement Learning

Zero-Shot LLMs in Human-in-the-Loop RL: Replacing Human Feedback for Reward Shaping

2025-03-26 · Mohammad Saif Nazir, Chayan Banerjee

Reinforcement learning often faces challenges with reward misalignment, where agents optimize for given rewards but fail to exhibit the desired behaviors. This occurs when the reward function incentivizes proxy behaviors…

continuous-controlContinuous Control

Murphys Laws of AI Alignment: Why the Gap Always Wins

2025-09-04 · Madhava Gaikwad arxiv

We study reinforcement learning from human feedback under misspecification. Sometimes human feedback is systematically wrong on certain types of inputs, like a broken compass that points the wrong way in specific regions…

Reinforcement Learning

Objective Decoupling in Social Reinforcement Learning: Recovering Ground Truth from Sycophantic Majorities

2026-02-08 · Majid Ghasemi, Mark Crowley arxiv

Contemporary AI alignment strategies rely on a fragile premise: that human feedback, while noisy, remains a fundamentally truthful signal. In this paper, we identify this assumption as Dogma 4 of Reinforcement Learning (…

Reinforcement Learning