Saved in:
Bibliographic Details
Main Authors: Abreu, Natalie, Zhang, Edwin, Malach, Eran, Saphra, Naomi
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.17669
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908500828880896
author Abreu, Natalie
Zhang, Edwin
Malach, Eran
Saphra, Naomi
author_facet Abreu, Natalie
Zhang, Edwin
Malach, Eran
Saphra, Naomi
contents Although language models are trained to mimic humans, the resulting systems display capabilities beyond the scope of any one person. To understand this phenomenon, we use a controlled setting to identify properties of the training data that lead a model to transcend the performance of its data sources. We build on previous work to outline three modes of transcendence, which we call skill denoising, skill selection, and skill generalization. We then introduce a knowledge graph-based setting in which simulated experts generate data based on their individual expertise. We highlight several aspects of data diversity that help to enable the model's transcendent capabilities. Additionally, our data generation setting offers a controlled testbed that we hope is valuable for future research in the area.
format Preprint
id arxiv_https___arxiv_org_abs_2508_17669
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Taxonomy of Transcendence
Abreu, Natalie
Zhang, Edwin
Malach, Eran
Saphra, Naomi
Artificial Intelligence
Although language models are trained to mimic humans, the resulting systems display capabilities beyond the scope of any one person. To understand this phenomenon, we use a controlled setting to identify properties of the training data that lead a model to transcend the performance of its data sources. We build on previous work to outline three modes of transcendence, which we call skill denoising, skill selection, and skill generalization. We then introduce a knowledge graph-based setting in which simulated experts generate data based on their individual expertise. We highlight several aspects of data diversity that help to enable the model's transcendent capabilities. Additionally, our data generation setting offers a controlled testbed that we hope is valuable for future research in the area.
title A Taxonomy of Transcendence
topic Artificial Intelligence
url https://arxiv.org/abs/2508.17669