Adaptive behaviors emerge from Hebbian-like plasticity through flexible attractors

Adaptive behavior is commonly framed as an optimization problem, in which organisms learn to maximize reward. Here we show that simple networks with local, Hebbian-like learning rules, which have no reward objective, nonetheless reproduce hallmark features of animal reward-seeking when placed in closed-loop environments where rewards arrive as ordinary sensory input. Six canonical local learning rules were tested under these conditions, and despite most rules being designed for unsupervised representation learning, all rules learned to seek out signal. However, at the parameters tested, only one rule, prospective configuration, robustly captured all of a set of tested phenomena: reward-magnitude-dependent learning rates, probability matching, memory retention across contexts, and switching between exploration and exploitation with heavy-tailed statistics. In these networks, actions that produce input become self-sustaining states (attractors) of network dynamics. Attractors reside in an activity-weight space and mediate behaviors, and adapt to their environments rather than being fixed properties of the networks. These flexible geometries provide a parsimonious account of how heavy-tailed runs of repeated choices, which resemble Levy flights seen in foraging, also change with the environment. A single framework then makes predictions about behavior and neural population geometries that can be used to investigate the learning dynamics of the brain.

Authors

Institutions

Publication Details

Journal
bioRxiv (Cold Spring Harbor Laboratory)
Published
2026-11-19
DOI
https://doi.org/10.64898/2026.11.18.767638
Primary Topic
Neural dynamics and brain function
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Adaptive behaviors emerge from Hebbian-like plasticity through flexible attractors

Jonah W. Brenner, Jeffrey C. Erlich, Chunyu A. Duan, Benjamin du Pont et al.
bioRxiv (Cold Spring Harbor Laboratory)
Neural dynamics and brain function
preprint

Adaptive behaviors emerge from Hebbian-like plasticity through flexible attractors

Jonah W. Brenner, Jeffrey C. Erlich, Chunyu A. Duan, Benjamin du Pont, Jason Chen, Chenguang Li, Joanna Aloor, Hanming Ye, Matas Kudarauskas
preprint en

Abstract

Adaptive behavior is commonly framed as an optimization problem, in which organisms learn to maximize reward. Here we show that simple networks with local, Hebbian-like learning rules, which have no reward objective, nonetheless reproduce hallmark features of animal reward-seeking when placed in closed-loop environments where rewards arrive as ordinary sensory input. Six canonical local learning rules were tested under these conditions, and despite most rules being designed for unsupervised representation learning, all rules learned to seek out signal. However, at the parameters tested, only one rule, prospective configuration, robustly captured all of a set of tested phenomena: reward-magnitude-dependent learning rates, probability matching, memory retention across contexts, and switching between exploration and exploitation with heavy-tailed statistics. In these networks, actions that produce input become self-sustaining states (attractors) of network dynamics. Attractors reside in an activity-weight space and mediate behaviors, and adapt to their environments rather than being fixed properties of the networks. These flexible geometries provide a parsimonious account of how heavy-tailed runs of repeated choices, which resemble Levy flights seen in foraging, also change with the environment. A single framework then makes predictions about behavior and neural population geometries that can be used to investigate the learning dynamics of the brain.

bioRxiv (Cold Spring Harbor Laboratory)
Brigham and Women's Hospital (US), Harvard University (US), Sainsbury Laboratory (GB), University of Pennsylvania (US), Stanford University (US)
Neural dynamics and brain function
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon