In-silico replications and hypothesis testing for model benchmarking

This satellite event will show how in-silico experimentation can be used as a practical framework for benchmarking computational models of the human brain. It builds directly on our recent efforts (Wang, Deb et al. in prep) that show how prior experiments can be replayed in silico and used to evaluate which findings replicate across candidate brain models and which do not. Participants will be invited to run in-silico experiments of their own design using a drag-and-drop executable modeling platform.

When
Sunday, May 17, 2026 12:45 – 2:15 pm
Where
Blue Heron

Add to calendar Google Calendar Outlook Apple & others (.ics)

Organizer

Ruolin Wang

Speakers

N Apurva Ratan Murty Nancy Kanwisher Antônio Mello Brad Duchaine Mayukh Deb Ruolin Wang

Program outline

Opening

Welcome and opening remarks
5 min
Research motivation and overview of the Cortex platform
5 min

How encoding models can move beyond prediction toward replication. Introducing our drag-and-drop executable modeling platform: Cortex.

Understanding encoding models as predictive tools
5 min

A brief introduction to how we build encoding models.

Session 1

Looking into the Past — Replicating Published Findings

From prediction to replication
5 min

An overview of representative in-silico tests we have already performed, and what insights they have revealed.

Hands-on Lab: revisiting published experiments
10–15 min

Participants use the Lab with curated stimuli from published studies to examine whether model predictions reproduce known experimental findings.

Group discussion and summary
10–15 min

A discussion of what the models captured, where they failed, and what replication successes and failures can reveal about both computational models and cognitive theories.

Session 2

Other Uses Beyond Replication — Education and scientific exploration

Invited talk: Teaching with interactive brain-model tools
5 min

How the platform can support teaching, demonstration, and conceptual understanding in vision science.

Invited talk: using encoding models to motivate and design new studies
5 min

Short perspectives on how forward-looking simulations and model-based hypotheses can complement replication work when planning new experiments and refining study designs.

Hands-on Lab: exploring new experimental ideas
10–15 min

Participants use the same Lab workflow with their own stimuli or modified examples to compare predicted responses across ROIs or models and explore how simulations might refine future experimental hypotheses.

Group discussion and summary
10–15 min

A discussion of how model simulations can help generate hypotheses, sharpen experimental designs, and reveal the limits of current models.

Closing

Closing remarks and next steps
5 min