Christopher Wang

*= equal contribution

Population Transformer: Learning Population-level Representations of Neural Activity

[ pdf ] [ code ]

tl;dr A spatio-temporal foundation model for intracranial brain data.

Geeling Chau* · Christopher Wang* · Sabera J Talukder · Vighnesh Subramaniam · Saraswati Soedarmadji · Yisong Yue · Boris Katz · Andrei Barbu in ICLR 2025 · Oral (top 2% of papers)

Brain Treebank: Large-scale intracranial recordings from naturalistic language stimuli

[ pdf ] [ website ]

tl;dr A dataset collected from patients watching 43 hours of movies while surgically implanted probes record their brain activity.

Christopher Wang* · Adam Uri Yaari* · Aaditya K Singh · Vighnesh Subramaniam · Dana Rosenfarb · Jan DeWitt · Pranav Misra · Joseph R. Madsen · Scellig Stone · Gabriel Kreiman · Boris Katz · Ignacio Cases · Andrei Barbu in NeurIPS 2024 · Oral (top 2% of papers)

BrainBits: How Much of the Brain are Generative Reconstruction Methods Using?

[ pdf ] [ code ]

tl;dr We measure how much diffusion models actually rely on the brain to do reconstruction. It might be less than you expect!

David Mayo* · Christopher Wang* · Asa Harbin* · Abdulrahman Alabdulkareem · Albert Eaton Shaw · Boris Katz · Andrei Barbu in NeurIPS 2024

Revealing Vision-Language Integration in the Brain with Multimodal Networks

[ pdf ]

tl;dr Multimodal models, e.g. CLIP, can be used to probe for multimodal processing in the brain.

Vighnesh Subramaniam · Colin Conwell · Christopher Wang · Gabriel Kreiman · Boris Katz · Ignacio Cases · Andrei Barbu in ICML 2024

BrainBERT: Self-supervised representation learning for intracranial recordings

[ pdf ] [ code ]

tl;dr A foundation model for single-channel intracranial brain recordings.

Christopher Wang · Vighnesh Subramaniam · Adam Uri Yaari · Gabriel Kreiman · Boris Katz · Ignacio Cases · Andrei Barbu in ICLR 2023

Learning a natural-language to LTL executable semantic parser for grounded robotics

[ pdf ] [ code ]

tl;dr We build a system that maps natural language commands → linear temporal logic → executions. During training, feedback comes from the final execution.

Christopher Wang · Candace Ross · Yen-Ling Kuo · Boris Katz · Andrei Barbu in CoRL 2020

ObjectNet: A large-scale bias-controlled dataset for pushing the limits of object recognition models

[ pdf ] [ website ]

tl;dr Many vision benchmarks for robots are sample biased due to the fact that they are scraped from the internet. We collect a test-set that corrects for correlations with background, angle, etc.

Andrei Barbu · David Mayo · Julian Alverio · William Luo · Christopher Wang · Dan Gutfreund · Josh Tenenbaum · Boris Katz in NeurIPS 2019