Beyond Sample-Level Feedback: Using Reference-Level Feedback to Guide Data Synthesis

• URL: http://arxiv.org/abs/2502.04511v3
• Date: Sat, 11 Oct 2025 16:30:00 GMT
• Title: Beyond Sample-Level Feedback: Using Reference-Level Feedback to Guide Data Synthesis
• Authors: Shuhaib Mehri, Xiusi Chen, Heng Ji, Dilek Hakkani-Tür,
• Abstract summary: We introduce Reference-Level Feedback, a paradigm that extracts desirable characteristics from carefully curated reference samples to guide the synthesis of higher-quality instruction-response pairs.<n>Experiments demonstrate that Reference-Level Feedback consistently outperforms traditional sample-level feedback methods, generalizes across model architectures, and produces high-quality and diverse data at low cost.
• Score: 54.15152681093108
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: High-quality instruction-tuning data is crucial for developing Large Language Models (LLMs) that can effectively navigate real-world tasks and follow human instructions. While synthetic data generation offers a scalable approach for creating such datasets, it imposes a quality ceiling where models trained on the data cannot outperform the LLM generating it. To overcome this limitation, we introduce Reference-Level Feedback, a paradigm that extracts desirable characteristics from carefully curated reference samples to guide the synthesis of higher-quality instruction-response pairs. Using this approach, we synthesize REFED, a dataset of 10K instruction-response pairs. Fine-tuning Llama-3.1-8B-Instruct and Mistral-7B-Instruct on REFED demonstrate state-of-the-art performance among similarly sized models, notably reaching a 43.96\% length-controlled win-rate on AlpacaEval 2.0. Extensive experiments demonstrate that Reference-Level Feedback consistently outperforms traditional sample-level feedback methods, generalizes across model architectures, and produces high-quality and diverse data at low cost.

Related papers

• BARE: Leveraging Base Language Models for Few-Shot Synthetic Data Generation [71.46236155101032]
  Current data generation methods rely on seed sets containing tens of thousands of examples to prompt instruction-tuned models.<n>We show that when working with only a few seed examples, instruction-tuned models produce insufficient diversity for downstream tasks.<n>We propose Base-Refine, a novel two-stage method that combines the diversity of base models with the quality assurance of instruction-tuned models.
  arXiv  Detail & Related papers  (2025-02-03T00:12:40Z)
• Few-Shot Optimized Framework for Hallucination Detection in Resource-Limited NLP Systems [1.0124625066746595]
  We introduce DeepSeek Few-shot optimization to enhance weak label generation through iterative prompt engineering.<n>We achieve high-quality annotations that considerably enhanced the performance of downstream models.<n>We further fine-tuned the Mistral-7B-Instruct-v0.3 model on these optimized annotations, enabling it to accurately detect hallucinations in resource-limited settings.
  arXiv  Detail & Related papers  (2025-01-28T01:26:22Z)
• Multi-Armed Bandit Approach for Optimizing Training on Synthetic Data [7.603659241572307]
  We propose a novel UCB-based training procedure combined with a dynamic usability metric. Our proposed metric integrates low-level and high-level information from synthetic images and their corresponding real and synthetic datasets. We show that our metric is an effective way to rank synthetic images based on their usability.
  arXiv  Detail & Related papers  (2024-12-06T23:36:36Z)
• Forewarned is Forearmed: Leveraging LLMs for Data Synthesis through Failure-Inducing Exploration [90.41908331897639]
  Large language models (LLMs) have significantly benefited from training on diverse, high-quality task-specific data. We present a novel approach, ReverseGen, designed to automatically generate effective training samples.
  arXiv  Detail & Related papers  (2024-10-22T06:43:28Z)
• Synthesizing Post-Training Data for LLMs through Multi-Agent Simulation [51.20656279478878]
  MATRIX is a multi-agent simulator that automatically generates diverse text-based scenarios.<n>We introduce MATRIX-Gen for controllable and highly realistic data synthesis.<n>On AlpacaEval 2 and Arena-Hard benchmarks, Llama-3-8B-Base, post-trained on datasets synthesized by MATRIX-Gen with just 20K instruction-response pairs, outperforms Meta's Llama-3-8B-Instruct model.
  arXiv  Detail & Related papers  (2024-10-18T08:01:39Z)
• Optimizing Instruction Synthesis: Effective Exploration of Evolutionary Space with Tree Search [25.108044778194536]
  We introduce IDEA-MCTS (Instruction Data Enhancement using Monte Carlo Tree Search), a scalable framework for efficiently synthesizing instructions. With tree search and evaluation models, it can efficiently guide each instruction to evolve into a high-quality form, aiding in instruction fine-tuning. Experimental results show that IDEA-MCTS significantly enhances the seed instruction data, raising the average evaluation scores of quality, diversity, and complexity from 2.19 to 3.81.
  arXiv  Detail & Related papers  (2024-10-14T11:28:30Z)
• Enhancing Unsupervised Sentence Embeddings via Knowledge-Driven Data Augmentation and Gaussian-Decayed Contrastive Learning [37.54523122932728]
  We propose a pipeline-based data augmentation method via large language models (LLMs) To tackle the issue of low data diversity, our pipeline utilizes knowledge graphs (KGs) to extract entities and quantities. To address high data noise, the GCSE model uses a Gaussian-decayed function to limit the impact of false hard negative samples.
  arXiv  Detail & Related papers  (2024-09-19T16:29:58Z)
• A Framework for Fine-Tuning LLMs using Heterogeneous Feedback [69.51729152929413]
  We present a framework for fine-tuning large language models (LLMs) using heterogeneous feedback. First, we combine the heterogeneous feedback data into a single supervision format, compatible with methods like SFT and RLHF. Next, given this unified feedback dataset, we extract a high-quality and diverse subset to obtain performance increases.
  arXiv  Detail & Related papers  (2024-08-05T23:20:32Z)
• Progressively Label Enhancement for Large Language Model Alignment [42.01694160556464]
  Large Language Models (LLM) alignment aims to prevent models from producing content that misaligns with human expectations. We propose PLE, a framework that dynamically adjusts the model's training process based on the evolving quality of the generated data.
  arXiv  Detail & Related papers  (2024-08-05T16:21:17Z)
• Unveiling the Flaws: Exploring Imperfections in Synthetic Data and Mitigation Strategies for Large Language Models [89.88010750772413]
  Synthetic data has been proposed as a solution to address the issue of high-quality data scarcity in the training of large language models (LLMs) Our work delves into these specific flaws associated with question-answer (Q-A) pairs, a prevalent type of synthetic data, and presents a method based on unlearning techniques to mitigate these flaws. Our work has yielded key insights into the effective use of synthetic data, aiming to promote more robust and efficient LLM training.
  arXiv  Detail & Related papers  (2024-06-18T08:38:59Z)
• DSF-GAN: DownStream Feedback Generative Adversarial Network [0.07083082555458872]
  We propose a novel architecture called the DownStream Feedback Generative Adversarial Network (DSF-GAN) DSF-GAN incorporates feedback from a downstream prediction model during training to augment the generator's loss function with valuable information. Our experiments demonstrate improved model performance when training on synthetic samples generated by DSF-GAN, compared to those generated by the same GAN architecture without feedback.
  arXiv  Detail & Related papers  (2024-03-27T05:41:50Z)
• RECOST: External Knowledge Guided Data-efficient Instruction Tuning [25.985023475991625]
  We argue that most current data-efficient instruction-tuning methods are highly dependent on the quality of the original instruction-tuning dataset. We propose a framework dubbed as textbfRECOST, which integrates external-knowledge-base re-ranking and diversity-consistent sampling into a single pipeline.
  arXiv  Detail & Related papers  (2024-02-27T09:47:36Z)
• Silkie: Preference Distillation for Large Visual Language Models [56.10697821410489]
  This paper explores preference distillation for large vision language models (LVLMs) We first build a vision-language feedback dataset utilizing AI annotation. We adopt GPT-4V to assess the generated outputs regarding helpfulness, visual faithfulness, and ethical considerations. The resulting model Silkie, achieves 6.9% and 9.5% relative improvement on the MME benchmark regarding the perception and cognition capabilities.
  arXiv  Detail & Related papers  (2023-12-17T09:44:27Z)
• TRIAGE: Characterizing and auditing training data for improved regression [80.11415390605215]
  We introduce TRIAGE, a novel data characterization framework tailored to regression tasks and compatible with a broad class of regressors. TRIAGE utilizes conformal predictive distributions to provide a model-agnostic scoring method, the TRIAGE score. We show that TRIAGE's characterization is consistent and highlight its utility to improve performance via data sculpting/filtering, in multiple regression settings.
  arXiv  Detail & Related papers  (2023-10-29T10:31:59Z)
• SURF: Semi-supervised Reward Learning with Data Augmentation for Feedback-efficient Preference-based Reinforcement Learning [168.89470249446023]
  We present SURF, a semi-supervised reward learning framework that utilizes a large amount of unlabeled samples with data augmentation. In order to leverage unlabeled samples for reward learning, we infer pseudo-labels of the unlabeled samples based on the confidence of the preference predictor. Our experiments demonstrate that our approach significantly improves the feedback-efficiency of the preference-based method on a variety of locomotion and robotic manipulation tasks.
  arXiv  Detail & Related papers  (2022-03-18T16:50:38Z)
• CAFE: Learning to Condense Dataset by Aligning Features [72.99394941348757]
  We propose a novel scheme to Condense dataset by Aligning FEatures (CAFE) At the heart of our approach is an effective strategy to align features from the real and synthetic data across various scales. We validate the proposed CAFE across various datasets, and demonstrate that it generally outperforms the state of the art.
  arXiv  Detail & Related papers  (2022-03-03T05:58:49Z)
• Abstractive Summarization for Low Resource Data using Domain Transfer and Data Synthesis [1.148539813252112]
  We explore using domain transfer and data synthesis to improve the performance of recent abstractive summarization methods. We show that tuning state of the art model trained on newspaper data could boost performance on student reflection data. We propose a template-based model to synthesize new data, which when incorporated into training further increased ROUGE scores.
  arXiv  Detail & Related papers  (2020-02-09T17:49:08Z)

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.