pen Life Lab
Back to reference library

Reference library

Vectorized instructive signals in cortical dendrites

Paper summary

Mice learned an experimenter-defined neurofeedback task using two intermingled retrosplenial layer-5 neuron populations with opposing decoder roles. Relative distal-versus-proximal calcium event magnitudes carried task information and opposing role-dependent signs. Surrounding-network activity predicted part of those residual signals. Perturbing dendritic processing through NDNF-positive layer-1 inhibition impaired learning, with light controls and decoder recalibration to address activity suppression.

Relevance to the project

Strong experimental motivation for considering neuron-specific teaching information separately from activity and global reward. It does not solve multiple receiver-objective scaling or conflict, identify our scorer interface, or authorize changing the adopted handover. The useful connection is a differentiated instruction component, not identity with our gradients.

Biological neuronsDendritic computationLearning and adaptation

Restricted mouse neurofeedback task with an externally known objective. No individual synaptic loss gradients or hierarchy-wide chain-rule computation were measured. Perturbation affected activity as well as instruction; feedback origin and plasticity versus online-computation contributions remain open. Proximal trunk is a somatic proxy, and negative residual means relative attenuation rather than negative calcium.

Source record

Reference ID
doi-10-1038-s41586-026-10190-7
Archived version
Published open-access article, Nature 652:1254–1263, online 25 February 2026; PDF includes Methods and extended data.
Summary provenance
reference-library/papers/doi-10-1038-s41586-026-10190-7/metadata.json · summary
Review status
Targeted source-grounded review; no experimental reproduction or exhaustive proof/figure audit
Reading scope
Published full-text PDF acquired. Main article, relevant main figure captions, Discussion and Methods on task/animals, imaging, optogenetics, decoder calibration, residual/event analyses and statistics read. Extended-data images, supplementary raw data and analysis code were not independently audited or reproduced.