Limited Preference Data? Learning Better Reward Model with Latent Space Synthesis

• URL: http://arxiv.org/abs/2509.26074v2
• Date: Tue, 14 Oct 2025 15:45:25 GMT
• Title: Limited Preference Data? Learning Better Reward Model with Latent Space Synthesis
• Authors: Leitian Tao, Xuefeng Du, Sharon Li,
• Abstract summary: Reward modeling is crucial for aligning large language models with human preferences.<n>Existing textual data synthesis methods are computationally expensive.<n>We propose a novel framework for synthesizing preference data directly in the language's latent embedding space.
• Score: 19.056892622506826
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reward modeling, crucial for aligning large language models (LLMs) with human preferences, is often bottlenecked by the high cost of preference data. Existing textual data synthesis methods are computationally expensive. We propose a novel framework LENS for synthesizing preference data directly in the LLM's latent embedding space. Our method employs a Variational Autoencoder (VAE) to learn a structured latent representation of response embeddings. By performing controlled perturbations in this latent space and decoding back to the embedding space, we efficiently generate diverse, semantically consistent synthetic preference pairs, bypassing costly text generation and annotation. We provide theoretical guarantees that our synthesized pairs approximately preserve original preference ordering and improve reward model generalization. Empirically, our latent-space synthesis significantly outperforms text-based augmentation on standard benchmarks, achieving superior results while being 18x faster in generation and using a 16,000x smaller model. Our work offers a scalable and effective alternative for enhancing reward modeling through efficient data augmentation. Code is publicly available at https://github.com/deeplearning-wisc/lens

Related papers

• Aligning Large Language Models via Fully Self-Synthetic Data [20.05693955243206]
  Traditional reinforcement learning from human feedback (RLHF) for large language models (LLMs) relies on expensive human-annotated datasets.<n>In this work, we introduce Self-Alignment Optimization (SAO), a fully self-synthetic framework for LLM alignment.<n>Experiments demonstrate that SAO effectively enhances the model's chat capabilities on standard benchmarks like AlpacaEval2.0.
  arXiv  Detail & Related papers  (2025-10-08T05:07:45Z)
• Scaling Transformer-Based Novel View Synthesis Models with Token Disentanglement and Synthetic Data [53.040873127309766]
  We propose a token disentanglement process within the transformer architecture, enhancing feature separation and ensuring more effective learning.<n>Our method outperforms existing models on both in-dataset and cross-dataset evaluations.
  arXiv  Detail & Related papers  (2025-09-08T17:58:06Z)
• Scaling Laws of Synthetic Data for Language Models [132.67350443447611]
  We introduce SynthLLM, a scalable framework that transforms pre-training corpora into diverse, high-quality synthetic datasets.<n>Our approach achieves this by automatically extracting and recombining high-level concepts across multiple documents using a graph algorithm.
  arXiv  Detail & Related papers  (2025-03-25T11:07:12Z)
• Synthetic Data Generation Using Large Language Models: Advances in Text and Code [0.0]
  Large language models (LLMs) are transforming synthetic training data generation in both natural language and code domains.<n>We highlight key techniques such as prompt-based generation, retrieval-augmented pipelines, and iterative self-refinement.<n>We discuss the accompanying challenges, including factual inaccuracies in generated text, insufficient stylistic or distributional realism, and risks of bias amplification.
  arXiv  Detail & Related papers  (2025-03-18T08:34:03Z)
• Synthetic Text Generation for Training Large Language Models via Gradient Matching [27.74603049449281]
  We propose the first theoretically rigorous approach for generating synthetic human-readable text.<n>We leverage Alternating Direction Method of Multipliers (ADMM) to iteratively optimize the embeddings of synthetic examples.<n>In doing so, the generated synthetic text guarantees convergence of the model to a close neighborhood of the solution obtained by fine-tuning on real data.
  arXiv  Detail & Related papers  (2025-02-24T19:49:15Z)
• MM-RLHF: The Next Step Forward in Multimodal LLM Alignment [59.536850459059856]
  We introduce MM-RLHF, a dataset containing $mathbf120k$ fine-grained, human-annotated preference comparison pairs.<n>We propose several key innovations to improve the quality of reward models and the efficiency of alignment algorithms.<n>Our approach is rigorously evaluated across $mathbf10$ distinct dimensions and $mathbf27$ benchmarks.
  arXiv  Detail & Related papers  (2025-02-14T18:59:51Z)
• DreamMask: Boosting Open-vocabulary Panoptic Segmentation with Synthetic Data [61.62554324594797]
  We propose DreamMask, which explores how to generate training data in the open-vocabulary setting, and how to train the model with both real and synthetic data.<n>In general, DreamMask significantly simplifies the collection of large-scale training data, serving as a plug-and-play enhancement for existing methods.<n>For instance, when trained on COCO and tested on ADE20K, the model equipped with DreamMask outperforms the previous state-of-the-art by a substantial margin of 2.1% mIoU.
  arXiv  Detail & Related papers  (2025-01-03T19:00:00Z)
• Little Giants: Synthesizing High-Quality Embedding Data at Scale [71.352883755806]
  We introduce SPEED, a framework that aligns open-source small models to efficiently generate large-scale embedding data. SPEED uses only less than 1/10 of the GPT API calls, outperforming the state-of-the-art embedding model E5_mistral when both are trained solely on their synthetic data.
  arXiv  Detail & Related papers  (2024-10-24T10:47:30Z)
• Private prediction for large-scale synthetic text generation [28.488459921169905]
  We present an approach for generating differentially private synthetic text using large language models (LLMs) In the private prediction framework, we only require the output synthetic data to satisfy differential privacy guarantees.
  arXiv  Detail & Related papers  (2024-07-16T18:28:40Z)
• PAL: Pluralistic Alignment Framework for Learning from Heterogeneous Preferences [6.398937923320069]
  We propose PAL, a framework to model human preference complementary to existing pretraining strategies. We show that PAL achieves competitive reward model accuracy compared to strong baselines.
  arXiv  Detail & Related papers  (2024-06-12T17:54:54Z)
• 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.<n>Our key idea is leveraging the human prior knowledge within the small (seed) data.<n>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)
• SynthesizRR: Generating Diverse Datasets with Retrieval Augmentation [55.2480439325792]
  We study the synthesis of six datasets, covering topic classification, sentiment analysis, tone detection, and humor. We find that SynthesizRR greatly improves lexical and semantic diversity, similarity to human-written text, and distillation performance.
  arXiv  Detail & Related papers  (2024-05-16T12:22:41Z)

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.