AIR: A Systematic Analysis of Annotations, Instructions, and Response Pairs in Preference Dataset

• URL: http://arxiv.org/abs/2504.03612v1
• Date: Fri, 04 Apr 2025 17:33:07 GMT
• Title: AIR: A Systematic Analysis of Annotations, Instructions, and Response Pairs in Preference Dataset
• Authors: Bingxiang He, Wenbin Zhang, Jiaxi Song, Cheng Qian, Zixuan Fu, Bowen Sun, Ning Ding, Haiwen Hong, Longtao Huang, Hui Xue, Ganqu Cui, Wanxiang Che, Zhiyuan Liu, Maosong Sun,
• Abstract summary: Preference learning is critical for aligning large language models with human values.<n>Our work shifts preference dataset design from ad hoc scaling to component-aware optimization.
• Score: 95.45316956434608
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Preference learning is critical for aligning large language models (LLMs) with human values, yet its success hinges on high-quality datasets comprising three core components: Preference \textbf{A}nnotations, \textbf{I}nstructions, and \textbf{R}esponse Pairs. Current approaches conflate these components, obscuring their individual impacts and hindering systematic optimization. In this work, we propose \textbf{AIR}, a component-wise analysis framework that systematically isolates and optimizes each component while evaluating their synergistic effects. Through rigorous experimentation, AIR reveals actionable principles: annotation simplicity (point-wise generative scoring), instruction inference stability (variance-based filtering across LLMs), and response pair quality (moderate margins + high absolute scores). When combined, these principles yield +5.3 average gains over baseline method, even with only 14k high-quality pairs. Our work shifts preference dataset design from ad hoc scaling to component-aware optimization, offering a blueprint for efficient, reproducible alignment.

Related papers

• Review, Refine, Repeat: Understanding Iterative Decoding of AI Agents with Dynamic Evaluation and Selection [71.92083784393418]
  Inference-time methods such as Best-of-N (BON) sampling offer a simple yet effective alternative to improve performance. We propose Iterative Agent Decoding (IAD) which combines iterative refinement with dynamic candidate evaluation and selection guided by a verifier.
  arXiv  Detail & Related papers  (2025-04-02T17:40:47Z)
• TD3: Tucker Decomposition Based Dataset Distillation Method for Sequential Recommendation [50.23504065567638]
  This paper introduces textbfTD3, a novel textbfDataset textbfDistillation method within a meta-learning framework.<n> TD3 distills a fully expressive emphsynthetic sequence summary from original data.<n>An augmentation technique allows the learner to closely fit the synthetic summary, ensuring an accurate update of it in the emphouter-loop.
  arXiv  Detail & Related papers  (2025-02-05T03:13:25Z)
• Iterative Feature Space Optimization through Incremental Adaptive Evaluation [30.404285897464256]
  Iterative feature space optimization involves systematically evaluating and adjusting the feature space to improve downstream task performance.<n>Existing works suffer from three key limitations.<n> overlooking differences among data samples leads to evaluation bias.<n> tailoring feature spaces to specific machine learning models results in overfitting and poor generalization.<n>We propose a gEneralized Adaptive feature Space Evaluator to efficiently produce optimal and generalized feature spaces.
  arXiv  Detail & Related papers  (2025-01-24T19:24:20Z)
• Anchored Preference Optimization and Contrastive Revisions: Addressing Underspecification in Alignment [57.03947082589616]
  Large Language Models (LLMs) are often aligned using contrastive alignment objectives and preference pair datasets. We study this and find that preference data gives a better learning signal when the underlying responses are contrastive. We introduce Contrastive Learning from AI Revisions (CLAIR), a data-creation method which leads to more contrastive preference pairs. Our best model, trained on 32K CLAIR preferences with APO, improves Llama-3-8B-Instruct by 7.65%, closing the gap with GPT4-turbo by 45%.
  arXiv  Detail & Related papers  (2024-08-12T16:24:51Z)
• Interpretable Triplet Importance for Personalized Ranking [5.409302364904161]
  We propose a shapely value-based method to measure the triplet importance in an interpretable manner. Our model consistently outperforms the state-of-the-art methods.
  arXiv  Detail & Related papers  (2024-07-28T11:46:55Z)
• Deep Content Understanding Toward Entity and Aspect Target Sentiment Analysis on Foundation Models [0.8602553195689513]
  Entity-Aspect Sentiment Triplet Extraction (EASTE) is a novel Aspect-Based Sentiment Analysis task. Our research aims to achieve high performance on the EASTE task and investigates the impact of model size, type, and adaptation techniques on task performance. Ultimately, we provide detailed insights and achieving state-of-the-art results in complex sentiment analysis.
  arXiv  Detail & Related papers  (2024-07-04T16:48:14Z)
• Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment [72.99676237703099]
  We propose a new framework that boosts the alignment of large language models with human preferences. Our key idea is leveraging the human prior knowledge within the small (seed) data. We introduce a noise-aware preference learning algorithm to mitigate the risk of low quality within generated preference data.
  arXiv  Detail & Related papers  (2024-06-06T18:01:02Z)
• Offline Regularised Reinforcement Learning for Large Language Models Alignment [33.483481840098925]
  We propose DRO, or emphDirect Reward optimisation, as a framework and associated algorithms. DRO uses a simple mean-squared objective that can be implemented in various ways.
  arXiv  Detail & Related papers  (2024-05-29T14:11:29Z)
• ROPO: Robust Preference Optimization for Large Language Models [59.10763211091664]
  We propose an iterative alignment approach that integrates noise-tolerance and filtering of noisy samples without the aid of external models. Experiments on three widely-used datasets with Mistral-7B and Llama-2-7B demonstrate that ROPO significantly outperforms existing preference alignment methods.
  arXiv  Detail & Related papers  (2024-04-05T13:58:51Z)
• Weakly-Supervised Aspect-Based Sentiment Analysis via Joint Aspect-Sentiment Topic Embedding [71.2260967797055]
  We propose a weakly-supervised approach for aspect-based sentiment analysis. We learn sentiment, aspect> joint topic embeddings in the word embedding space. We then use neural models to generalize the word-level discriminative information.
  arXiv  Detail & Related papers  (2020-10-13T21:33:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.