What Has a Foundation Model Found? Using Inductive Bias to Probe for World Models

Keyon Vafa, Peter G. Chang, Ashesh Rambachan, Sendhil Mullainathan
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:60727-60747, 2025.

Abstract

Foundation models are premised on the idea that sequence prediction can uncover deeper domain understanding, much like how Kepler’s predictions of planetary motion later led to the discovery of Newtonian mechanics. However, evaluating whether these models truly capture deeper structure remains a challenge. We develop a technique for evaluating foundation models that examines how they adapt to synthetic datasets generated from some postulated world model. Our technique measures whether the foundation model’s inductive bias aligns with the world model, and so we refer to it as an inductive bias probe. Across multiple domains, we find that foundation models can excel at their training tasks yet fail to develop inductive biases towards the underlying world model when adapted to new tasks. We particularly find that foundation models trained on orbital trajectories consistently fail to apply Newtonian mechanics when adapted to new physics tasks. Further analysis reveals that these models behave as if they develop task-specific heuristics that fail to generalize.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-vafa25a, title = {What Has a Foundation Model Found? {U}sing Inductive Bias to Probe for World Models}, author = {Vafa, Keyon and Chang, Peter G. and Rambachan, Ashesh and Mullainathan, Sendhil}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {60727--60747}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/vafa25a/vafa25a.pdf}, url = {https://proceedings.mlr.press/v267/vafa25a.html}, abstract = {Foundation models are premised on the idea that sequence prediction can uncover deeper domain understanding, much like how Kepler’s predictions of planetary motion later led to the discovery of Newtonian mechanics. However, evaluating whether these models truly capture deeper structure remains a challenge. We develop a technique for evaluating foundation models that examines how they adapt to synthetic datasets generated from some postulated world model. Our technique measures whether the foundation model’s inductive bias aligns with the world model, and so we refer to it as an inductive bias probe. Across multiple domains, we find that foundation models can excel at their training tasks yet fail to develop inductive biases towards the underlying world model when adapted to new tasks. We particularly find that foundation models trained on orbital trajectories consistently fail to apply Newtonian mechanics when adapted to new physics tasks. Further analysis reveals that these models behave as if they develop task-specific heuristics that fail to generalize.} }
Endnote
%0 Conference Paper %T What Has a Foundation Model Found? Using Inductive Bias to Probe for World Models %A Keyon Vafa %A Peter G. Chang %A Ashesh Rambachan %A Sendhil Mullainathan %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-vafa25a %I PMLR %P 60727--60747 %U https://proceedings.mlr.press/v267/vafa25a.html %V 267 %X Foundation models are premised on the idea that sequence prediction can uncover deeper domain understanding, much like how Kepler’s predictions of planetary motion later led to the discovery of Newtonian mechanics. However, evaluating whether these models truly capture deeper structure remains a challenge. We develop a technique for evaluating foundation models that examines how they adapt to synthetic datasets generated from some postulated world model. Our technique measures whether the foundation model’s inductive bias aligns with the world model, and so we refer to it as an inductive bias probe. Across multiple domains, we find that foundation models can excel at their training tasks yet fail to develop inductive biases towards the underlying world model when adapted to new tasks. We particularly find that foundation models trained on orbital trajectories consistently fail to apply Newtonian mechanics when adapted to new physics tasks. Further analysis reveals that these models behave as if they develop task-specific heuristics that fail to generalize.
APA
Vafa, K., Chang, P.G., Rambachan, A. & Mullainathan, S.. (2025). What Has a Foundation Model Found? Using Inductive Bias to Probe for World Models. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:60727-60747 Available from https://proceedings.mlr.press/v267/vafa25a.html.

Related Material