JEPA · 2026-10-05
NeuroLens: Learning Latent Embeddings of Neural Semantics from Chronic Recordings
Hanrui Lyu, Baiyuan Chen, Tianshu Tan, Matthew R. Whiteway, Maxwell D. Melin, Ji Xia, Linyang He, Bradly C. Stadie, Anne Churchland, Liam Paninski, Yizi Zhang
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
NeuroLens is a self-supervised JEPA-style model for chronic neural recordings that learns denoised, semantically informative latent embeddings. A day-adaptive cross-attention encoder maps changing neuron populations into a common latent space using calibration-based neuron identification, and a shared causal transformer predicts future latent states. Predicting in latent space reduces sensitivity to recording instability and improves decoding over time.
Why it matters
Chronic neural recordings mix representational plasticity with recording nonstationarity (e.g., drift and neuron turnover). The authors propose a general framework that aims to separate task-relevant structure from recording-specific variability by learning predictive latent embeddings. They report improved decoding of decision-making and semantic task variables, generalization to future sessions after multi-day pretraining, and more stable decoding over time than state-of-the-art baselines.
Method
- Day-adaptive encoder: cross-attention aggregates neuron tokens in a common latent space while using calibration trials to infer neuron identities across days (via firing-rate statistics, autocorrelograms, and population coupling).
- Shared causal temporal predictor: a transformer shared across days predicts future latent embeddings using only preceding latents (strictly causal).
- Training objective: latent prediction MSE plus SIGReg regularization to prevent representation collapse (encouraging nonzero variance across latent dimensions).
Abstract (from arXiv)
Understanding how neural activity represents higher-order cognition and how these representations evolve over time has long been a central pursuit in neuroscience. However, current analytical tools cannot easily distinguish representational plasticity from recording instability in chronic neural recordings. Here, we introduce NeuroLens (Latent Embeddings of Neural Semantics), a self-supervised model based on the Joint-Embedding Predictive Architecture (JEPA) framework that learns denoised, semantically informative latents from chronic neural recordings. An adaptive encoder maps changing neural populations into a common latent space, while a temporal predictor learns structure that supports prediction of future latent states. By predicting in latent space, NeuroLens captures temporally predictable structure and reduces sensitivity to transient, recording-specific variability. Across chronic intracortical data in mice and humans, the learned representations improve decoding of decision-making and semantic task variables. Multi-day pretraining enables generalization to future sessions, rapid few-shot adaptation to unseen neural populations, and more stable decoding over time than state-of-the-art baselines. Together, these results establish NeuroLens as a new paradigm for studying how neural representations change during learning and over long timescales.
Related papers
- Latent Prediction Yields Cross-Subject Alignable Representations in Sensory-Evoked fMRI
- World Model Interpretability: From Decoding Representations to Understanding Latent Worlds
- TEX-MEX: Similarity-Guided Temporal Contrastive Learning for Time Series
- AC-JEPA: Anchored Cross-Modal Joint Embedding Predictive Architecture for Visual-to-Symbolic Translation
- Learning Intrinsic Dynamics in Latent World Models
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-02 · Indexed by Pub-AI: 2026-10-05 UTC
NeuroLens: Learning Latent Embeddings of Neural Semantics from Chronic Recordings
Preprint posted: 2026-10-02 · Original record ↗
- arxiv
- 2610.02864
Explicit source versions (1)
- v1 · 2026-10-02
- Preprint available
Source date: 2026-10-02 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Understanding how neural activity represents higher-order cognition and how these representations evolve over time has long been a central pursuit in neuroscience. However, current analytical tools cannot easily distinguish representational plasticity from recording instability in chronic neural recordings. Here, we introduce NeuroLens (Latent Embeddings of Neural Semantics), a self-supervised model based on the Joint-Embedding Predictive Architecture (JEPA) framework that learns denoised, semantically informative latents from chronic neural recordings. An adaptive encoder maps changing neural populations into a common latent space, while a temporal predictor learns structure that supports prediction of future latent states. By predicting in latent space, NeuroLens captures temporally predictable structure and reduces sensitivity to transient, recording-specific variability. Across chronic intracortical data in mice and humans, the learned representations improve decoding of decision-making and semantic task variables. Multi-day pretraining enables generalization to future sessions, rapid few-shot adaptation to unseen neural populations, and more stable decoding over time than state-of-the-art baselines. Together, these results establish NeuroLens as a new paradigm for studying how neural representations change during learning and over long timescales.