---
pretty_name: Multimodal Brain Scaling
license: mit
tags:
- neuroscience
- neuroai
- computer-vision
- scaling-laws
- brain-alignment
- visual-cortex
- fmri
- eeg
- meg
- electrophysiology
configs:
- config_name: pretraining
  data_files:
  - split: results
    path: data/pretraining_results_with_metadata.parquet
- config_name: layer_search
  data_files:
  - split: results
    path: data/layer_search_results.parquet
- config_name: finetuning
  data_files:
  - split: results
    path: data/finetuning_results.parquet
- config_name: mapping
  data_files:
  - split: results
    path: data/mapping_results.parquet
---

# Multimodal Scaling Laws for Task & Data-Optimized Models of Visual Cortex

[![Project Page](https://img.shields.io/badge/Project%20Page-EPFL%20site-E60028.svg?logo=googlechrome&logoColor=white)](https://multimodal-brain-scaling.epfl.ch)
[![GitHub](https://img.shields.io/badge/GitHub-Code-181717.svg?logo=github&logoColor=white)](https://github.com/epflneuroailab/multimodal-brain-scaling)
[![OpenReview](https://img.shields.io/badge/OpenReview-OQ6jQHJPTT-1A3D91.svg?logoColor=white)](https://openreview.net/forum?id=OQ6jQHJPTT)
[![ICML 2026](https://img.shields.io/badge/ICML-2026-0B5D1E.svg?logoColor=white)](https://icml.cc/virtual/2026/poster/64356)

This repository hosts the result tables and accompanying metadata released
with our **ICML 2026** paper
*Multimodal Scaling Laws for Task & Data-Optimized Models of Visual Cortex*.

## Paper summary

Task-optimized deep networks are the leading in-silico models of sensory
cortex, but progress is fragmented across datasets, modalities, and
evaluation protocols, making it hard to identify which modeling choices
actually drive better brain alignment. We address this gap with a unified,
large-scale benchmark of model-to-brain alignment spanning **8 neural
datasets** (macaque electrophysiology and human fMRI / EEG / MEG) and
**>600 vision models** evaluated under a consistent pipeline. Along three
scaling axes — pretraining resources, neural fine-tuning, and the mapping
from features to neural responses — we report three robust trends:

1. **Pretraining saturation.** Alignment improves with pretraining
   compute and data, but saturates at large scale across every recording
   modality.
2. **Complementary neural fine-tuning.** Hybrid task + neural objectives
   yield consistent, transferable gains over pure task pretraining.
3. **Mapping scaling.** Increasing paired stimulus--response data for the
   mapping yields log-linear gains that dominate alignment improvements
   in many regimes.

We also introduce a lightweight, subject-shared cross-attention readout
that matches or surpasses per-subject linear decoders with an order of
magnitude fewer parameters. Together, these results provide practical
guidance for allocating data, compute, and supervision when building
next-generation brain-aligned models.

## Repository contents

Large result tables are stored as Parquet files under `data/`; small
companion files (dataset dimensions, normalization constants, model
metadata) live under `metadata/`.

### Result tables (`data/`)

- `pretraining`: Pretraining-scale model-to-brain alignment results with model metadata.
- `layer_search`: All-layer search results used to choose model layers for benchmark evaluation.
- `finetuning`: Neural fine-tuning and LoRA result table.
- `mapping`: Mapping-data scaling and readout comparison result table.

### Metadata (`metadata/`)

- `dataset_dims.json`: per-dataset stimulus counts, subjects, and ROI sizes.
- `model_metadata.csv` / `model_metadata_compact.csv`: model family,
  architecture, pretraining data, parameter count, and training compute
  for every backbone evaluated.
- `nc_summary.csv`: per-ROI noise ceilings used to normalize alignment
  scores.
- `mapping_params_per_roi.csv`: parameter budgets of the mapping models
  per ROI.

## Quick start

```python
from datasets import load_dataset

# Main pretraining-scale alignment results with model metadata.
ds = load_dataset("epfl-neuroai/multimodal-brain-scaling", "pretraining", split="results")
df = ds.to_pandas()
```

Other splits follow the same pattern (`layer_search`, `finetuning`,
`mapping`).

To rehydrate the local CSV artifacts expected by the analysis code in the
[GitHub repository](https://github.com/epflneuroailab/multimodal-brain-scaling):

```bash
python -m mbs.download_artifacts
```

## Citation

```bibtex
@inproceedings{
  gokce2026multimodal_brain_scaling,
  title     = {Multimodal Scaling Laws for Task \& Data-Optimized Models of Visual Cortex},
  author    = {Abdulkadir Gokce and Yingtian Tang and Martin Schrimpf},
  booktitle = {Forty-third International Conference on Machine Learning},
  year      = {2026},
  url       = {https://openreview.net/forum?id=OQ6jQHJPTT}
}
```

## License

Released under the MIT License. Underlying neural datasets retain their
original licenses; please refer to the respective sources before
redistribution.