Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating World

• URL: http://arxiv.org/abs/2410.16713v1
• Date: Tue, 22 Oct 2024 05:49:24 GMT
• Title: Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating World
• Authors: Joshua Kazdan, Rylan Schaeffer, Apratim Dey, Matthias Gerstgrasser, Rafael Rafailov, David L. Donoho, Sanmi Koyejo,
• Abstract summary: We study collapse versus avoidance of collapse when generative machine learning models are pretrained on web-scale datasets. Surprisingly, we find a non-trivial interaction between real and synthetic data, where the value of synthetic data for reducing test loss depends on the absolute quantity of real data.
• Score: 19.266191284270793
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: The increasing presence of AI-generated content on the internet raises a critical question: What happens when generative machine learning models are pretrained on web-scale datasets containing data created by earlier models? Some authors prophesy $\textit{model collapse}$ under a "$\textit{replace}$" scenario: a sequence of models, the first trained with real data and each later one trained only on synthetic data from its preceding model. In this scenario, models successively degrade. Others see collapse as easily avoidable; in an "$\textit{accumulate}$' scenario, a sequence of models is trained, but each training uses all real and synthetic data generated so far. In this work, we deepen and extend the study of these contrasting scenarios. First, collapse versus avoidance of collapse is studied by comparing the replace and accumulate scenarios on each of three prominent generative modeling settings; we find the same contrast emerges in all three settings. Second, we study a compromise scenario; the available data remains the same as in the accumulate scenario -- but unlike $\textit{accumulate}$ and like $\textit{replace}$, each model is trained using a fixed compute budget; we demonstrate that model test loss on real data is larger than in the $\textit{accumulate}$ scenario, but apparently plateaus, unlike the divergence seen with $\textit{replace}$. Third, we study the relative importance of cardinality and proportion of real data for avoiding model collapse. Surprisingly, we find a non-trivial interaction between real and synthetic data, where the value of synthetic data for reducing test loss depends on the absolute quantity of real data. Our insights are particularly important when forecasting whether future frontier generative models will collapse or thrive, and our results open avenues for empirically and mathematically studying the context-dependent value of synthetic data.

Related papers

• A Theoretical Perspective: How to Prevent Model Collapse in Self-consuming Training Loops [55.07063067759609]
  High-quality data is essential for training large generative models, yet the vast reservoir of real data available online has become nearly depleted. Models increasingly generate their own data for further training, forming Self-consuming Training Loops (STLs) Some models degrade or even collapse, while others successfully avoid these failures, leaving a significant gap in theoretical understanding.
  arXiv  Detail & Related papers  (2025-02-26T06:18:13Z)
• How to Synthesize Text Data without Model Collapse? [37.219627817995054]
  Model collapse in synthetic data indicates that iterative training on self-generated data leads to a gradual decline in performance. We propose token editing on human-produced data to obtain semi-synthetic data.
  arXiv  Detail & Related papers  (2024-12-19T09:43:39Z)
• Universality of the $π^2/6$ Pathway in Avoiding Model Collapse [0.0]
  We demonstrate the universality of the pi-squared-over-6 augment risk bound across a large family of canonical statistical models. We provide a framework that is able to accommodate a large variety of augment processes.
  arXiv  Detail & Related papers  (2024-10-30T08:44:10Z)
• Self-Consuming Generative Models with Curated Data Provably Optimize Human Preferences [20.629333587044012]
  We study the impact of data curation on iterated retraining of generative models. We prove that, if the data is curated according to a reward model, the expected reward of the iterative retraining procedure is maximized.
  arXiv  Detail & Related papers  (2024-06-12T21:28:28Z)
• Beyond Model Collapse: Scaling Up with Synthesized Data Requires Verification [11.6055501181235]
  We investigate the use of verification on synthesized data to prevent model collapse. We show that verifiers, even imperfect ones, can indeed be harnessed to prevent model collapse.
  arXiv  Detail & Related papers  (2024-06-11T17:46:16Z)
• How Bad is Training on Synthetic Data? A Statistical Analysis of Language Model Collapse [9.59833542807268]
  Model collapse occurs when new models are trained on synthetic data generated from previously trained models. We show that model collapse cannot be avoided when training solely on synthetic data. We estimate a maximal amount of synthetic data below which model collapse can eventually be avoided.
  arXiv  Detail & Related papers  (2024-04-07T22:15:13Z)
• Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data [49.73114504515852]
  We show that replacing the original real data by each generation's synthetic data does indeed tend towards model collapse. We demonstrate that accumulating the successive generations of synthetic data alongside the original real data avoids model collapse.
  arXiv  Detail & Related papers  (2024-04-01T18:31:24Z)
• Towards Theoretical Understandings of Self-Consuming Generative Models [56.84592466204185]
  This paper tackles the emerging challenge of training generative models within a self-consuming loop. We construct a theoretical framework to rigorously evaluate how this training procedure impacts the data distributions learned by future models. We present results for kernel density estimation, delivering nuanced insights such as the impact of mixed data training on error propagation.
  arXiv  Detail & Related papers  (2024-02-19T02:08:09Z)
• Learning Defect Prediction from Unrealistic Data [57.53586547895278]
  Pretrained models of code have become popular choices for code understanding and generation tasks. Such models tend to be large and require commensurate volumes of training data. It has become popular to train models with far larger but less realistic datasets, such as functions with artificially injected bugs. Models trained on such data tend to only perform well on similar data, while underperforming on real world programs.
  arXiv  Detail & Related papers  (2023-11-02T01:51:43Z)
• On the Stability of Iterative Retraining of Generative Models on their own Data [56.153542044045224]
  We study the impact of training generative models on mixed datasets. We first prove the stability of iterative training under the condition that the initial generative models approximate the data distribution well enough. We empirically validate our theory on both synthetic and natural images by iteratively training normalizing flows and state-of-the-art diffusion models.
  arXiv  Detail & Related papers  (2023-09-30T16:41:04Z)
• 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)
• Datamodels: Predicting Predictions from Training Data [86.66720175866415]
  We present a conceptual framework, datamodeling, for analyzing the behavior of a model class in terms of the training data. We show that even simple linear datamodels can successfully predict model outputs.
  arXiv  Detail & Related papers  (2022-02-01T18:15:24Z)
• Variational Bayesian Unlearning [54.26984662139516]
  We study the problem of approximately unlearning a Bayesian model from a small subset of the training data to be erased. We show that it is equivalent to minimizing an evidence upper bound which trades off between fully unlearning from erased data vs. not entirely forgetting the posterior belief. In model training with VI, only an approximate (instead of exact) posterior belief given the full data can be obtained, which makes unlearning even more challenging.
  arXiv  Detail & Related papers  (2020-10-24T11:53:00Z)
• STAN: Synthetic Network Traffic Generation with Generative Neural Models [10.54843182184416]
  This paper presents STAN (Synthetic network Traffic generation with Autoregressive Neural models), a tool to generate realistic synthetic network traffic datasets. Our novel neural architecture captures both temporal dependencies and dependence between attributes at any given time. We evaluate the performance of STAN in terms of the quality of data generated, by training it on both a simulated dataset and a real network traffic data set.
  arXiv  Detail & Related papers  (2020-09-27T04:20:02Z)

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.