SAN: Hypothesizing Long-Term Synaptic Development and Neural Engram Mechanism in Scalable Model's Parameter-Efficient Fine-Tuning

Gaole Dai, Chun-Kai Fan, Yiming Tang, Zhi Zhang, Yuan Zhang, Yulu Gan, Qizhe Zhang, Cheng-Ching Tseng, Shanghang Zhang, Tiejun Huang

Abstract:

Advances in Parameter-efficient Fine-tuning (PEFT) bridged the performance gap with Full Fine-Tuning (FFT) through sophisticated analysis of pre-trained parameter spaces. Starting from drawing insights from Neural Engrams (NE) in Biological Neural Networks (BNNs), we establish a connection between the low-rank property observed during PEFT's parameter space shifting and neurobiological mechanisms. This observation leads to our proposed method, **S**ynapse and **N**euron (**SAN**), which decomposes and propagates the scaling component from anterior feature adjustment vectors towards posterior weight matrices. Our approach is theoretically grounded in Long-Term Potentiation/Depression (LTP/D) phenomena, which govern synapse development through neurotransmitter release modulation.
Extensive experiments demonstrate its effectiveness: on **vision tasks** across VTAB, FGVC, and GIC (25 datasets) using ViT, Swin-T and ConvNeXt architectures, SAN outperforms FFT up to *8.7%* and LoRA by *3.2%*; on **language tasks** using Commonsense Reasoning (8 datasets) with LLaMA models (all generations), surpassing ChatGPT up to *8.5%* and LoRA by *4.7%*; on **vision-language tasks** using Visual Instruction Tuning (7 datasets) with LLaVA models, it exceeds FFT up to *2.4%* and LoRA by *1.9%*. Our code and W&B log will be released

Lay Summary:

How can we teach massive AI models new skills without the huge cost and effort of retraining them entirely? Current efficient methods help, but we looked to the human brain for a smarter way.

Our research introduces a method called Synapse and Neuron (SAN). It's inspired by how our brains efficiently learn by strengthening or weakening connections between neurons—a process linked to how memories form. SAN mimics this by observing early adjustments as an AI learns a new task. It then intelligently passes on a "scaling" signal to later parts of the model, preparing them effectively without adding new trainable components.

Remarkably, SAN significantly boosted performance on diverse tasks—analyzing images, understanding language, and even combined visual-language challenges. It outperformed traditional full retraining by up to 8.7% and another popular efficient technique, LoRA, by up to 4.7% across various benchmarks. Our brain-inspired "plug-and-play" approach offers a more efficient and powerful path for adapting large AI models to new challenges.

Primary Area: Deep Learning->Everything Else

Keywords: Parameter-efficient Fine-tuning, Foundation Models, Neuroscience, Neural Engrams, Long-term Synaptic Development

Submission Number: 2153