Format
3-hour introductory tutorial
Affiliated Event
AI4BCI 2026 related tutorial
Tutorial
Language Models as Windows into Human Language Processing: Neural Encoding, Decoding, and Brain Alignment with NLP
This affiliated tutorial introduces neural encoding and neural decoding to the NLP community through a brain-language modeling perspective. It asks how NLP can help us understand language processing in the human brain, and how brain data can help evaluate and improve language models.
The tutorial presents neural encoding as a way to test whether representations from language models, including contextual embeddings, surprisal, attention patterns, and layer-wise features, can explain brain responses during reading or listening.
It also introduces neural decoding as the reverse problem: reconstructing linguistic or multimodal content from brain activity using retrieval, representation learning, and foundation-model-guided generation, while emphasizing uncertainty, limitations, consent, and mental privacy.
Audience
NLP, machine learning, and cognitive science researchers
Background
No prior neuroscience background required
Estimated Size
50-75 participants
Tutorial outline
Three-hour structure
- 10 mins
Opening: Why brain alignment matters for NLP
Encoding versus decoding, the essentials of fMRI, EEG, MEG, and ECoG for NLP researchers, and the tutorial roadmap.
- 85 mins
Part 1: Neural encoding and bidirectional brain alignment
Foundations of encoding models, LLM features for understanding the brain, brain-informed model comparison, multilingual and cross-subject alignment, and a minimal encoding-model demo.
- 75 mins
Part 2: Brain decoding as foundation-model-guided generation
Decoding regimes, LLM priors for language decoding, vision-language models for multimodal decoding, semantic evaluation, failure analysis, and robustness checks.
- 10 mins
Synthesis, ethics, and classroom transfer
Shared limitations, mental privacy, informed consent, benchmark design, and a reusable classroom exercise for graduate NLP or cognitive science courses.
People
Presenters and support
Presenter
Dr. Jingyuan Sun
Department of Computer Science, University of Manchester
Show bioHide bio
Dr. Jingyuan Sun is an Assistant Professor in the Department of Computer Science at the University of Manchester. His research lies at the intersection of artificial intelligence, computational linguistics, and cognitive neuroscience, with a focus on multimodal neural encoding and decoding from non-invasive brain recordings.
Presenter
Dr. Yunhao Zhang
Institute of Automation, Chinese Academy of Sciences
Show bioHide bio
Dr. Yunhao Zhang is a Research Assistant Professor at the Institute of Automation, Chinese Academy of Sciences. His research connects artificial intelligence, cognitive neuroscience, and psycholinguistics, using large language models to analyze multimodal neuroimaging data and study semantic representation in the human brain.
Tutoring assistant
Yifan Wang
Department of Computer Science, University of Manchester
Tutoring assistant
Yang Cui
Department of Computer Science, University of Manchester
Audience
Who it is for
The tutorial is designed for graduate students, postdoctoral researchers, faculty, and practitioners interested in LLM interpretability, human-like language understanding, multimodal learning, assistive communication, and brain-language modeling.
A basic understanding of machine learning and neural networks is helpful. Familiarity with Transformers and contextual embeddings will be useful, and Python knowledge can help attendees follow the demos, but none of these are required for participation.
Slides, a curated bibliography, notebook-style demos, and reusable teaching materials are planned for public release where dataset licenses permit.
Brain alignmentNeural encodingNeural decodingLarge language modelsComputational neuroscienceHuman language processingfMRIEEGMEGECoGBrain-computer interfacesModel interpretabilitySemantic representationMultimodal generation
Back to Affiliated Events