NLP
CORE MACHINE LEARNING
What is my math transformer doing? Three results on interpretability and generalization
December 02, 2022
Abstract
This paper investigates the failure cases and out-of-distribution behavior of trans- formers trained on matrix inversion and eigenvalue decomposition. I show that incorrect model predictions still retain deep mathematical properties of the solution (e.g. correct eigenvalues, unit norm of eigenvectors), and that almost all model fail- ures can be attributed to, and predicted from, properties of the problem or solution. This demonstrates that, when in doubt, math transformers do not hallucinate absurd solutions (as was sometimes proposed) but remain “roughly right”. I also show that the careful choice of a training dataset can accelerate training, while allowing the model to generalize out of its training distribution, invalidating the idea that transformers “merely interpolate” from memorized examples.
AUTHORS
Written by
François Charton
Publisher
Neurips MAH-AI workshop
Related Publications
June 05, 2026
CONVERSATIONAL AI
RANKING AND RECOMMENDATIONS
Superintelligent Retrieval Agent: The Next Frontier of Agentic Retrieval
Anshumali Shrivastava, Jason Chen, Qi Ma, Zeyu Yang
June 05, 2026
May 20, 2026
HUMAN & MACHINE INTELLIGENCE
RESEARCH
EgoBabyVLM: Benchmarking Cross-Modal Learning from Naturalistic Egocentric Video Data
Alvin W. M. Tan, Nicolas Hamilakis, Manel Khentout, Sho Tsuji, Balázs Kégl, Michael C. Frank, Angel Villar Corrales, Charles-Eric Saint-James, Dongyan Lin, Emmanuel Dupoux, Jiayi Shen, Juan Pino, Mahi Luthra, Martin Gleize, Phillip Rust, Rashel Moritz, Sheila Krogh-Jespersen, Surya Parimi, Tom Fizycki, Vanessa Stark, Yosuke Higuchi, Youssef Benchekroun
May 20, 2026
May 18, 2026
CONVERSATIONAL AI
RESEARCH
GIM: Evaluating models via tasks that integrate multiple cognitive domains
Alexandre Rezende, Rohit Patel, Steven McClain
May 18, 2026
May 12, 2026
HUMAN & MACHINE INTELLIGENCE
RESEARCH
NeuralSet: A High-Performing Python Package for Neuro-AI
Corentin Bel, Linnea Evanson, Julien Gadonneix, Andrea Santos Revilla, Mingfang (Lucy) Zhang, Julie Bonnaire, Charlotte Caucheteux, Alexandre Défossez, Théo Desbordes, Pablo Diego-Simón, Shubh Khanna, Juliette Millet, Pierre Orhan, Saarang Panchavati, Antoine Ratouchniak, Alexis Thual, Hubert Jacob Banville, Jarod Levy, Jean Remi King, Josephine Raugel, Jérémy Rapin, Katelyn Begany, Marlene Careil, Simon Dahan, Sophia Houhamdi, Stéphane d'Ascoli, Teon Brooks, Yohann Benchetrit
May 12, 2026
Help Us Pioneer The Future of AI
We share our open source frameworks, tools, libraries, and models for everything from research exploration to large-scale production deployment.
Our approach
Latest news
Foundational models
Meta © 2026