ARC Prize presents

ARC Prize Research Summit 2026

Boston, Massachusetts · October 23, 2026

Frontier AI systems are increasingly neither purely deep learning nor purely symbolic. The most promising advances are emerging at the intersection of these paradigms, where neural networks, program synthesis, and formal reasoning can combine in novel ways.

The ARC Prize Research Summit is a forum for researchers advancing this frontier.

Hosted by the ARC Prize Foundation in collaboration with MIT.

Speaker lineup

Stephen Wolfram

Wolfram Research

Founder of Wolfram Research and creator of Mathematica and Wolfram|Alpha. His book A New Kind of Science investigates the complex behavior that can emerge from simple computational rules.

Kevin Murphy

Google DeepMind

Active learning of causal world models: the Model Discovery Agent combines probabilistic inference with language models to propose and learn causal explanations from data, guiding the experiments needed to test them.

Alexis Fox

Duke University

Lead author of PRO-LONG, a minimal framework that gives AI agents programmatic access to their full interaction history. The work explores how programmatic memory supports long-horizon reasoning while reducing token costs on ARC-AGI-3.

Alexandre Bouayad

G-Research / INRIA

Weave of Formal Thought (WoFT) connects formal syntax with learned representations. The work combines constrained decoding and latent-variable fine-tuning to help language models use grammar as a structural scratchpad for reasoning.

François Chollet

ARC Prize Foundation / Ndea

François Chollet gave the opening remarks at the summit. He created Keras and the ARC-AGI benchmark and co-founded the ARC Prize Foundation and Ndea.

Summit speakers and organizers posing for a group photo
Attendees listening to a presentation at the summit
A presenter pointing to an ARC-AGI-3 research poster during a discussion