Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2602.06852 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910014337187840 |
|---|---|
| author | Pan, Jonathan |
| author_facet | Pan, Jonathan |
| contents | Mechanistic interpretability aims to reverse-engineer the internal computations of Large Language Models (LLMs), yet separating sparse semantic signals from high-dimensional polysemantic noise remains a significant challenge. This paper introduces the Quantum Sieve Tracer, a hybrid quantum-classical framework designed to characterize factual recall circuits. We implement a modular pipeline that first localizes critical layers using classical causal tracing, then maps specific attention head activations into an exponentially large quantum Hilbert space. Using open-weight models (Meta Llama-3.2-1B and Alibaba Qwen2.5-1.5B-Instruct), we perform a two-stage analysis that reveals a fundamental architectural divergence. While Qwen's layer 7 circuit functions as a classic Recall Hub, we discover that Llama's layer 9 acts as an Interference Suppression circuit, where ablating the identified heads paradoxically improves factual recall. Our results demonstrate that quantum kernels can distinguish between these constructive (recall) and reductive (suppression) mechanisms, offering a high-resolution tool for analyzing the fine-grained topology of attention. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_06852 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | The Quantum Sieve Tracer: A Hybrid Framework for Layer-Wise Activation Tracing in Large Language Models Pan, Jonathan Quantum Physics Artificial Intelligence Mechanistic interpretability aims to reverse-engineer the internal computations of Large Language Models (LLMs), yet separating sparse semantic signals from high-dimensional polysemantic noise remains a significant challenge. This paper introduces the Quantum Sieve Tracer, a hybrid quantum-classical framework designed to characterize factual recall circuits. We implement a modular pipeline that first localizes critical layers using classical causal tracing, then maps specific attention head activations into an exponentially large quantum Hilbert space. Using open-weight models (Meta Llama-3.2-1B and Alibaba Qwen2.5-1.5B-Instruct), we perform a two-stage analysis that reveals a fundamental architectural divergence. While Qwen's layer 7 circuit functions as a classic Recall Hub, we discover that Llama's layer 9 acts as an Interference Suppression circuit, where ablating the identified heads paradoxically improves factual recall. Our results demonstrate that quantum kernels can distinguish between these constructive (recall) and reductive (suppression) mechanisms, offering a high-resolution tool for analyzing the fine-grained topology of attention. |
| title | The Quantum Sieve Tracer: A Hybrid Framework for Layer-Wise Activation Tracing in Large Language Models |
| topic | Quantum Physics Artificial Intelligence |
| url | https://arxiv.org/abs/2602.06852 |