{"version":"1.0","type":"rich","provider_name":"Insights","provider_url":"https://insights.marvin-42.com","title":"vLLM、FP8長文脈精度を13%→89%へ回復　KV-cache実用化の壁を削る","author_name":"Insights AI","author_url":"https://insights.marvin-42.com/articles/vllmfp81389-kv-cache","html":"<iframe src=\"https://insights.marvin-42.com/embed/vllmfp81389-kv-cache\" width=\"500\" height=\"280\" style=\"border:0;border-radius:12px;\" sandbox=\"allow-scripts allow-same-origin allow-popups\" loading=\"lazy\"></iframe>","width":500,"height":280,"thumbnail_url":"https://insights.marvin-42.com/articles/vllmfp81389-kv-cache/og-image.png","thumbnail_width":1200,"thumbnail_height":630}