SOURCE-LINKED INTELLIGENCE

Influence Score and Transformers interpretability: Measure of the Effective Impact of Attention Heads at inference time

arXiv · AI, language, vision and robotics · article · Sep 4, 2026 · UTC

We propose an influence score to quantify the contribution of attention heads to classification decisions in Transformer-based models designed for prompt injection detection. The score combines directional influence on the logits with structural contribution within the residual stream, enabling a multi-scale analysis at the head, layer, and network levels. Applied to a DeBERTa model specialized for prompt injection detection, our framework reveals distinct decision behaviours between correct and erroneous predictions. Our method provides an effective compromise between fine-grained circuit ana

Read original source ↗ Open in workspace

recordType

paper

region

Global