arxiv
PublishedJune 11, 2026 at 4:00 AM
The Structural Attention Tax: How Retrieval Format Hijacks In-Context Learning Independent of Content
Publisher summary· verbatim
arXiv:2606.11198v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) systems inject external knowledge to improve LLM outputs, yet the format of injected content -- distinct from its semantic relevance -- can independently distort the model's attention distribution. We identify and
Stay posted· Newsletter
A 5-min weekly brief — top movers, price watch, story of the week.
Discussion
No replies yet. Be first.
Related coverage
More from ARXIV
arxivMODF-SIR: A Multi-agent Omni-modal Distilled Framework for Social Intelligence Reasoning17harxivPosition: Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces!17harxivARGUS: Stacked Multi-View Identity Mosaic Injection for Subject-Preserving Video Generation17harxivGeneralizing Beyond Suboptimality: Offline Reinforcement Learning Learns Effective Scheduling through Random Solutions17hThe Bubble Brief
WEEKLYRead AI insights every Tuesday — top movers, new releases, story of the week.
Originally published on arxiv ↗