Neuron-Aware Active Few-Shot Learning for LLMs

2026-07-02

AI Summary

This work addresses a key limitation in existing active few-shot learning methods, which predominantly rely on output-layer signals—such as prediction entropy—to select samples and thus struggle to accurately identify knowledge blind spots in large language models. To overcome this, the authors propose NeuFS, a novel framework that, for the first time, incorporates internal neuron activation patterns into active sample selection. NeuFS dynamically represents samples through neuron-level activations and introduces a dual-criterion strategy that jointly optimizes for diversity coverage and neuron consensus, effectively pinpointing challenging instances prone to hallucination. Experimental results across three benchmark datasets demonstrate that NeuFS significantly outperforms current baselines in both reasoning and text classification tasks, validating the efficacy and novelty of leveraging internal neural signals to enhance few-shot learning performance.

Abstract

Active Few-Shot Learning (AFSL) adapts LLMs to specialized domains by identifying the most valuable unlabeled samples for annotation and use as few-shot demonstrations, effectively reducing human annotation costs while promoting high performance. However, existing methods typically rely on output-level signals for sample identification, such as predictive entropy or semantic similarities with test-time data based on external embeddings, which often overlook models' internal dynamics, which could pinpoint specific knowledge gaps. To bridge this gap, we propose NeuFS, a Neuron-Aware Active Few-Shot Learning framework that shifts the selection paradigm from output-level proxies to models' internal dynamics. NeuFS utilizes neuron activation patterns to represent sample directly, and includes a dual-criteria selection strategy that: (1) ensures few-shot sample diversity with neuron patterns for broader example coverage, while (2) prioritizing on identifying informative and challenging few-shot samples LLMs tend to hallucinate by quantifying neuron consensus. Experiments on three datasets demonstrate that NeuFS excels in both reasoning and text classification tasks, outperforming existing AFSL baselines. Ablation studies further highlight that internal neuron activations provide a more principled and effective selection signal than external embeddings, validating the superiority of the proposed NeuFS.

Problem

• Active Few-Shot Learning
• Large Language Models
• Neuron Activation
• Sample Selection
• Internal Dynamics

Innovation

• Neuron-Aware
• Active Few-Shot Learning
• Neuron Activation
• Sample Selection
• Large Language Models