<?xml version="1.0" encoding="UTF-8"?><urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/attention-sinks</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/the-discovery-why-first-tokens-hoard-attention</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/why-removing-initial-tokens-breaks-generation</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/the-streamingllm-solution</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/mathematical-analysis</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/worked-example-tracing-one-eviction-step</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/implementation</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/designing-effective-sink-tokens</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/streaming-inference-pipeline</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/empirical-validation</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/beyond-streamingllm-learnable-sink-tokens</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/in-practice-deploying-streaming-inference</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/key-parameters</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/limitations-and-considerations</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/summary</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url><url><loc>https://mbrenndoerfer.com/writing/attention-sinks-streamingllm-infinite-generation/quiz</loc><lastmod>2025-07-04T00:00:00.000Z</lastmod></url></urlset>