{"ID":2851640,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.19875","arxiv_id":"2510.19875","title":"Stream: Scaling up Mechanistic Interpretability to Long Context in LLMs via Sparse Attention","abstract":"As Large Language Models (LLMs) scale to million-token contexts, traditional Mechanistic Interpretability techniques for analyzing attention scale quadratically with context length, demanding terabytes of memory beyond 100,000 tokens. We introduce Sparse Tracing, a novel technique that leverages dynamic sparse attention to efficiently analyze long context attention patterns. We present Stream, a compilable hierarchical pruning algorithm that estimates per-head sparse attention masks in near-linear time $O(T \\log T)$ and linear space $O(T)$, enabling one-pass interpretability at scale. Stream performs a binary-search-style refinement to retain only the top-$k$ key blocks per query while preserving the model's next-token behavior. We apply Stream to long chain-of-thought reasoning traces and identify thought anchors while pruning 97-99\\% of token interactions. On the RULER benchmark, Stream preserves critical retrieval paths while discarding 90-96\\% of interactions and exposes layer-wise routes from the needle to output. Our method offers a practical drop-in tool for analyzing attention patterns and tracing information flow without terabytes of caches. By making long context interpretability feasible on consumer GPUs, Sparse Tracing helps democratize chain-of-thought monitoring. Code is available at https://anonymous.4open.science/r/stream-03B8/.","short_abstract":"As Large Language Models (LLMs) scale to million-token contexts, traditional Mechanistic Interpretability techniques for analyzing attention scale quadratically with context length, demanding terabytes of memory beyond 100,000 tokens. We introduce Sparse Tracing, a novel technique that leverages dynamic sparse attentio...","url_abs":"https://arxiv.org/abs/2510.19875","url_pdf":"https://arxiv.org/pdf/2510.19875v2","authors":"[\"J Rosser\",\"José Luis Redondo García\",\"Gustavo Penha\",\"Konstantina Palla\",\"Hugues Bouchard\"]","published":"2025-10-22T09:42:29Z","proceeding":"cs.CL","tasks":"[\"cs.CL\",\"cs.AI\"]","methods":"[\"Large Language Model\",\"Language Model\"]","project_urls":"[\"https://anonymous.4open.science/r/stream-03B8/\"]","has_code":false}