Neural Preference Learning (NPL) is a novel architecture that gives LLM agents persistent, personal preferences by pairing them with a companion spiking neural network. Unlike RLHF (which is batch, pre-deployment, and population-level), NPL operates in real-time, learning from individual user feedback through natural language.
reinforcement-learning hebbian-learning preference-learning lif-neuron spiking-neural-network rlhf llm-agent openclaw brainjar
-
Updated
Mar 16, 2026 - JavaScript