Transcendence: Generative Models Can Outperform The Experts That Train Them

• URL: http://arxiv.org/abs/2406.11741v4
• Date: Sat, 12 Oct 2024 18:46:20 GMT
• Title: Transcendence: Generative Models Can Outperform The Experts That Train Them
• Authors: Edwin Zhang, Vincent Zhu, Naomi Saphra, Anat Kleiman, Benjamin L. Edelman, Milind Tambe, Sham M. Kakade, Eran Malach,
• Abstract summary: We study the phenomenon of transcendence: when a generative model achieves capabilities that surpass the abilities of the experts generating its data. We demonstrate transcendence by training an autoregressive transformer to play chess from game transcripts, and show that the trained model can sometimes achieve better performance than all players in the dataset.
• Score: 55.885802048647655
• License:
• Abstract: Generative models are trained with the simple objective of imitating the conditional probability distribution induced by the data they are trained on. Therefore, when trained on data generated by humans, we may not expect the artificial model to outperform the humans on their original objectives. In this work, we study the phenomenon of transcendence: when a generative model achieves capabilities that surpass the abilities of the experts generating its data. We demonstrate transcendence by training an autoregressive transformer to play chess from game transcripts, and show that the trained model can sometimes achieve better performance than all players in the dataset. We theoretically prove that transcendence can be enabled by low-temperature sampling, and rigorously assess this claim experimentally. Finally, we discuss other sources of transcendence, laying the groundwork for future investigation of this phenomenon in a broader setting.

Related papers

• Will the Inclusion of Generated Data Amplify Bias Across Generations in Future Image Classification Models? [29.71939692883025]
  We investigate the effects of generated data on image classification tasks, with a specific focus on bias. Hundreds of experiments are conducted on Colorized MNIST, CIFAR-20/100, and Hard ImageNet datasets. Our findings contribute to the ongoing debate on the implications of synthetic data for fairness in real-world applications.
  arXiv  Detail & Related papers  (2024-10-14T05:07:06Z)
• 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)
• Humanoid Locomotion as Next Token Prediction [84.21335675130021]
  Our model is a causal transformer trained via autoregressive prediction of sensorimotor trajectories. We show that our model enables a full-sized humanoid to walk in San Francisco zero-shot. Our model can transfer to the real world even when trained on only 27 hours of walking data, and can generalize commands not seen during training like walking backward.
  arXiv  Detail & Related papers  (2024-02-29T18:57:37Z)
• Fantastic Gains and Where to Find Them: On the Existence and Prospect of General Knowledge Transfer between Any Pretrained Model [74.62272538148245]
  We show that for arbitrary pairings of pretrained models, one model extracts significant data context unavailable in the other. We investigate if it is possible to transfer such "complementary" knowledge from one model to another without performance degradation.
  arXiv  Detail & Related papers  (2023-10-26T17:59:46Z)
• 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)
• Learning to Jump: Thinning and Thickening Latent Counts for Generative Modeling [69.60713300418467]
  Learning to jump is a general recipe for generative modeling of various types of data. We demonstrate when learning to jump is expected to perform comparably to learning to denoise, and when it is expected to perform better.
  arXiv  Detail & Related papers  (2023-05-28T05:38:28Z)
• When does Bias Transfer in Transfer Learning? [89.22641454588278]
  Using transfer learning to adapt a pre-trained "source model" to a downstream "target task" can dramatically increase performance with seemingly no downside. We demonstrate that there can exist a downside after all: bias transfer, or the tendency for biases of the source model to persist even after adapting the model to the target class.
  arXiv  Detail & Related papers  (2022-07-06T17:58:07Z)
• Adversarial Training Helps Transfer Learning via Better Representations [17.497590668804055]
  Transfer learning aims to leverage models pre-trained on source data to efficiently adapt to target setting. Recent works empirically demonstrate that adversarial training in the source data can improve the ability of models to transfer to new domains. We show that adversarial training in the source data generates provably better representations, so fine-tuning on top of this representation leads to a more accurate predictor of the target data.
  arXiv  Detail & Related papers  (2021-06-18T15:41:07Z)
• Residual Energy-Based Models for Text [46.22375671394882]
  We show that the generations of auto-regressive language models can be reliably distinguished from real text by statistical discriminators. This suggests that the auto-regressive models can be improved by incorporating the (globally normalized) discriminators into the generative process.
  arXiv  Detail & Related papers  (2020-04-06T13:44:03Z)

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.