An Information-Theoretic Criterion for Efficient Data Synthesis

• URL: http://arxiv.org/abs/2605.16379v1
• Date: Mon, 11 May 2026 01:27:59 GMT
• Title: An Information-Theoretic Criterion for Efficient Data Synthesis
• Abstract summary: Synthetic data is crucial for large language model training, but its effectiveness is highly inconsistent.<n>We provide an information-theoretic account of this inconsistency.
• Score: 11.130200422847162
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Synthetic data becomes crucial for large language model training, but its effectiveness is highly inconsistent. We provide an information-theoretic account of this inconsistency: synthetic data improves a model only when the generation-training loop is information-open, i.e., shaped by external signals (verifiers, environments, or rubrics) that inject task-relevant information beyond the model's current distribution. When the loop is information-closed (relying on the model's own outputs without such signals), the data processing inequality ensures that task-relevant information can only decrease, making collapse a predicted outcome. Among information-open pipelines, both efficiency and generalization hinge on the meta-level of supervision: a coarser signal such as binary correctness treats all acceptable outputs as equivalent, so the behavior it teaches is not tied to any particular domain or surface form and generalizes naturally across tasks and domains. These observations lead to a guiding thesis: learning preferentially converges to the most information-efficient signal component available, which accelerates learning when that component is the intended one, but causes reward hacking when a spurious pattern happens to be simpler.

Related papers

• Hidden Threat in Synthetic Data: Covert Targeted Bias Injection through Benign Text [9.060748681063144]
  We investigate whether social biases can be injected into aligned models through semantically benign synthetic data.<n>Experiments show that benign-looking synthetic data can act as a covert channel for transmitting targeted biases.
  arXiv  Detail & Related papers  (2026-08-31T11:28:36Z)
• From Entropy to Epiplexity: Rethinking Information for Computationally Bounded Intelligence [91.54446789584826]
  Epiplexity is a formalization of information capturing what computationally bounded observers can learn from data.<n>We show how information can be created with computation, how it depends on the ordering of the data, and how likelihood modeling can produce more complex programs than present in the data generating process itself.
  arXiv  Detail & Related papers  (2026-01-06T18:04:03Z)
• A Differentiable Adversarial Framework for Task-Aware Data Subsampling [0.5371337604556311]
  We introduce the antagonistic soft selection subsampling (ASSS) framework as a novel paradigm that reconstructs data reduction into a differentiable end-to-end learning problem.<n>This work establishes task aware data subsampling as a learnable component, providing a principled solution for effective large-scale data learning.
  arXiv  Detail & Related papers  (2026-01-05T13:10:09Z)
• Data coarse graining can improve model performance [7.325551965751601]
  We study a paradox using a solvable model of high-dimensional, ridge-regularized linear regression under 'data coarse graining'<n>Inspired by the renormalization group in statistical physics, we analyze coarse-graining schemes that systematically discard features based on their relevance to the learning task.<n>Our results highlight a complex, nonmonotonic risk landscape shaped by the structure of the data, and illustrate how ideas from statistical physics provide a principled lens for understanding modern machine learning phenomena.
  arXiv  Detail & Related papers  (2025-09-18T00:17:01Z)
• Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating World [19.266191284270793]
  generative machine learning models are pretrained on web-scale datasets containing data generated by earlier models.<n>Some prior work warns of "model collapse" as the web is overwhelmed by synthetic data.<n>We report experiments on three ways of using data (training-workflows) across three generative model task-settings.
  arXiv  Detail & Related papers  (2024-10-22T05:49:24Z)
• Impact of Noisy Supervision in Foundation Model Learning [91.56591923244943]
  This paper is the first work to comprehensively understand and analyze the nature of noise in pre-training datasets.<n>We propose a tuning method (NMTune) to affine the feature space to mitigate the malignant effect of noise and improve generalization.
  arXiv  Detail & Related papers  (2024-03-11T16:22:41Z)
• From Fake to Real: Pretraining on Balanced Synthetic Images to Prevent Spurious Correlations in Image Recognition [64.59093444558549]
  We propose a simple, easy-to-implement, two-step training pipeline that we call From Fake to Real. By training on real and synthetic data separately, FFR does not expose the model to the statistical differences between real and synthetic data. Our experiments show that FFR improves worst group accuracy over the state-of-the-art by up to 20% over three datasets.
  arXiv  Detail & Related papers  (2023-08-08T19:52:28Z)
• Synthetic data, real errors: how (not) to publish and use synthetic data [86.65594304109567]
  We show how the generative process affects the downstream ML task. We introduce Deep Generative Ensemble (DGE) to approximate the posterior distribution over the generative process model parameters.
  arXiv  Detail & Related papers  (2023-05-16T07:30:29Z)
• Boosting Differentiable Causal Discovery via Adaptive Sample Reweighting [62.23057729112182]
  Differentiable score-based causal discovery methods learn a directed acyclic graph from observational data. We propose a model-agnostic framework to boost causal discovery performance by dynamically learning the adaptive weights for the Reweighted Score function, ReScore.
  arXiv  Detail & Related papers  (2023-03-06T14:49:59Z)
• On-the-fly Denoising for Data Augmentation in Natural Language Understanding [101.46848743193358]
  We propose an on-the-fly denoising technique for data augmentation that learns from soft augmented labels provided by an organic teacher model trained on the cleaner original data. Our method can be applied to general augmentation techniques and consistently improve the performance on both text classification and question-answering tasks.
  arXiv  Detail & Related papers  (2022-12-20T18:58:33Z)
• Annotating and Modeling Fine-grained Factuality in Summarization [36.88018450067003]
  A major barrier to their use in practice is their propensity to output summaries that are not faithful to the input and that contain factual errors. We explore both synthetic and human-labeled data sources for training models to identify factual errors in summarization. We show that our best factuality detection model enables training of more factual XSum summarization models by allowing us to identify non-factual tokens in the training data.
  arXiv  Detail & Related papers  (2021-04-09T11:20:44Z)
• Negative Data Augmentation [127.28042046152954]
  We show that negative data augmentation samples provide information on the support of the data distribution. We introduce a new GAN training objective where we use NDA as an additional source of synthetic data for the discriminator. Empirically, models trained with our method achieve improved conditional/unconditional image generation along with improved anomaly detection capabilities.
  arXiv  Detail & Related papers  (2021-02-09T20:28:35Z)

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.