{"id":47164,"date":"2026-01-31T01:10:00","date_gmt":"2026-01-30T18:10:00","guid":{"rendered":"https:\/\/thaipropertynews.com\/feeds\/?p=47164"},"modified":"2026-01-31T01:10:00","modified_gmt":"2026-01-30T18:10:00","slug":"zilliz-open-sources-industry-first-bilingual-semantic-highlighting-model-to-slash-rag-token-costs-and-boost-accuracy","status":"publish","type":"post","link":"https:\/\/thaipropertynews.com\/feeds\/?p=47164","title":{"rendered":"Zilliz Open Sources Industry-First Bilingual &#8220;Semantic Highlighting&#8221; Model to Slash RAG Token Costs and Boost Accuracy"},"content":{"rendered":"<table border=\"0\" cellspacing=\"10\" cellpadding=\"5\" align=\"right\">\n<tbody>\n<tr>\n<td><img decoding=\"async\" src=\"https:\/\/mma.prnasia.com\/media2\/2713871\/zilliz_milvus_Logo.jpg?p=medium600\" border=\"0\" alt=\"\" title=\"logo\" hspace=\"0\" vspace=\"0\" width=\"118\" \/><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><span class=\"legendSpanClass\"><span class=\"xn-location\">REDWOOD CITY, Calif.<\/span>, Jan. 31, 2026 \/PRNewswire\/ &#8212; <u><a href=\"https:\/\/zilliz.com\/cloud?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Zilliz<\/a><\/u>, the company behind the leading open-source vector database <u><a href=\"https:\/\/milvus.io\/?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Milvus<\/a><\/u>, today announced the open-source release of its <u><a href=\"https:\/\/huggingface.co\/zilliz\/semantic-highlight-bilingual-v1\" target=\"_blank\" rel=\"nofollow\">Bilingual Semantic Highlighting Model<\/a><\/u>, an industry-first AI model designed to dramatically reduce <span>token<\/span> usage and improve answer quality in production RAG-powered AI applications.<\/span><\/p>\n<p>This highlighting model introduces sentence-level relevance filtering, enabling AI developers to remove low-signal context before sending prompts to large language models. This approach directly addresses rising inference costs and accuracy issues caused by oversized context windows in enterprise RAG and RAG-powered AI deployments.<\/p>\n<p>&#8220;As RAG systems move into production, teams are running into very real cost and quality limits,&#8221; said <span class=\"xn-person\">James Luan<\/span>, VP of Engineering at Zilliz. &#8220;This model gives developers a practical way to reduce prompt size and improve answer accuracy without reworking their existing pipelines.&#8221;<\/p>\n<div class=\"PRN_ImbeddedAssetReference\">\n<p><a href=\"https:\/\/mma.prnasia.com\/media2\/2873657\/Traditional_Highlight_VS_Semantic_Highlight.html\" target=\"_blank\" rel=\"nofollow\"><img decoding=\"async\" src=\"https:\/\/mma.prnasia.com\/media2\/2873657\/Traditional_Highlight_VS_Semantic_Highlight.jpg?p=medium600\" title=\"Traditional Highlight VS Semantic Highlight\" alt=\"Traditional Highlight VS Semantic Highlight\" \/><\/a><br \/><span>Traditional Highlight VS Semantic Highlight<\/span><\/p>\n<\/div>\n<p><b>Key Innovations and Technical Breakthroughs<\/b><\/p>\n<ul type=\"disc\">\n<li><b>Bilingual relevance by design:<\/b> Optimized for both English and Chinese, the model addresses cross-lingual relevance challenges common in global RAG deployments. It is built on the MiniCPM-2B architecture, enabling low-latency, production-ready performance.\n<\/li>\n<li><b>Sentence-level context filtering: <\/b>Rather than scoring entire document chunks, the model evaluates relevance at the sentence level and retains only content that directly supports a user query before sending it to the LLM.\n<\/li>\n<li><b>Lower <span>token<\/span> usage, higher answer quality: <\/b>Zilliz reports that sentence-level filtering significantly compresses prompt size while improving downstream response quality, helping teams reduce inference costs and improve generation speed in production environments.<\/li>\n<\/ul>\n<p><b>Availability<\/b><\/p>\n<p>The Bilingual Semantic Highlighting Model is available today as an open-source release. To learn more about the training methodology and performance benchmarks, visit the <u><a href=\"https:\/\/milvus.io\/blog\/semantic-highlighting-model-for-rag-context-pruning-and-token-saving.md?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Zilliz Technical Blog<\/a><\/u>.<\/p>\n<p>Download: : <u><a href=\"https:\/\/huggingface.co\/zilliz\/semantic-highlight-bilingual-v1\" target=\"_blank\" rel=\"nofollow\">zilliz\/semantic-highlight-bilingual-v1<\/a><\/u><\/p>\n<p><b>About Zilliz<\/b><\/p>\n<p>Zilliz is the company behind <u><a href=\"https:\/\/milvus.io\/?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Milvus<\/a><\/u>, the world&#8217;s most widely adopted open-source vector database. <u><a href=\"https:\/\/zilliz.com\/cloud?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Zilliz Cloud<\/a><\/u> brings that performance to production with a fully managed, cloud-native platform built for scalable, low-latency vector search and hybrid retrieval. It supports billion-scale workloads with sub-10ms latency, auto-scaling, and optimized indexes for GenAI use cases like semantic search and RAG.<\/p>\n<p>Zilliz is built to make AI not just possible\u2014but practical. With a focus on performance and cost-efficiency, it helps engineering teams move from prototype to production without overprovisioning or complex infrastructure. Over 10,000 organizations worldwide rely on Zilliz to build intelligent applications at scale.<\/p>\n<p>Headquartered in Redwood Shores, <span class=\"xn-location\">California<\/span>, Zilliz is backed by leading investors, including Aramco&#8217;s Prosperity 7 Ventures, Temasek&#8217;s Pavilion Capital, Hillhouse Capital, 5Y Capital, Yunqi Partners, Trustbridge Partners, and others. Learn more at\u00a0 <u><a href=\"https:\/\/zilliz.com\/?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Zilliz.com<\/a><\/u>.<\/p>\n<p>\u00a0<\/p>\n<div class=\"PRN_ImbeddedAssetReference\">\n<\/div>","protected":false},"excerpt":{"rendered":"<p><!-- wp:html --><\/p>\n<table border=\"0\" cellspacing=\"10\" cellpadding=\"5\" align=\"right\">\n<tbody>\n<tr>\n<td><img decoding=\"async\" src=\"https:\/\/mma.prnasia.com\/media2\/2713871\/zilliz_milvus_Logo.jpg?p=medium600\" border=\"0\" alt=\"\" title=\"logo\" hspace=\"0\" vspace=\"0\" width=\"118\" \/><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><span class=\"legendSpanClass\"><span class=\"xn-location\">REDWOOD CITY, Calif.<\/span>, Jan. 31, 2026 \/PRNewswire\/ &#8212; <u><a href=\"https:\/\/zilliz.com\/cloud?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Zilliz<\/a><\/u>, the company behind the leading open-source vector database <u><a href=\"https:\/\/milvus.io\/?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Milvus<\/a><\/u>, today announced the open-source release of its <u><a href=\"https:\/\/huggingface.co\/zilliz\/semantic-highlight-bilingual-v1\" target=\"_blank\" rel=\"nofollow\">Bilingual Semantic Highlighting Model<\/a><\/u>, an industry-first AI model designed to dramatically reduce <span>token<\/span> usage and improve answer quality in production RAG-powered AI applications.<\/span><\/p>\n<p>This highlighting model introduces sentence-level relevance filtering, enabling AI developers to remove low-signal context before sending prompts to large language models. This approach directly addresses rising inference costs and accuracy issues caused by oversized context windows in enterprise RAG and RAG-powered AI deployments.<\/p>\n<p>&#8220;As RAG systems move into production, teams are running into very real cost and quality limits,&#8221; said <span class=\"xn-person\">James Luan<\/span>, VP of Engineering at Zilliz. &#8220;This model gives developers a practical way to reduce prompt size and improve answer accuracy without reworking their existing pipelines.&#8221;<\/p>\n<div class=\"PRN_ImbeddedAssetReference\">\n<p><a href=\"https:\/\/mma.prnasia.com\/media2\/2873657\/Traditional_Highlight_VS_Semantic_Highlight.html\" target=\"_blank\" rel=\"nofollow\"><img decoding=\"async\" src=\"https:\/\/mma.prnasia.com\/media2\/2873657\/Traditional_Highlight_VS_Semantic_Highlight.jpg?p=medium600\" title=\"Traditional Highlight VS Semantic Highlight\" alt=\"Traditional Highlight VS Semantic Highlight\" \/><\/a><br \/><span>Traditional Highlight VS Semantic Highlight<\/span><\/p>\n<\/div>\n<p><b>Key Innovations and Technical Breakthroughs<\/b><\/p>\n<ul type=\"disc\">\n<li><b>Bilingual relevance by design:<\/b> Optimized for both English and Chinese, the model addresses cross-lingual relevance challenges common in global RAG deployments. It is built on the MiniCPM-2B architecture, enabling low-latency, production-ready performance.\n<\/li>\n<li><b>Sentence-level context filtering: <\/b>Rather than scoring entire document chunks, the model evaluates relevance at the sentence level and retains only content that directly supports a user query before sending it to the LLM.\n<\/li>\n<li><b>Lower <span>token<\/span> usage, higher answer quality: <\/b>Zilliz reports that sentence-level filtering significantly compresses prompt size while improving downstream response quality, helping teams reduce inference costs and improve generation speed in production environments.<\/li>\n<\/ul>\n<p><b>Availability<\/b><\/p>\n<p>The Bilingual Semantic Highlighting Model is available today as an open-source release. To learn more about the training methodology and performance benchmarks, visit the <u><a href=\"https:\/\/milvus.io\/blog\/semantic-highlighting-model-for-rag-context-pruning-and-token-saving.md?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Zilliz Technical Blog<\/a><\/u>.<\/p>\n<p>Download: : <u><a href=\"https:\/\/huggingface.co\/zilliz\/semantic-highlight-bilingual-v1\" target=\"_blank\" rel=\"nofollow\">zilliz\/semantic-highlight-bilingual-v1<\/a><\/u><\/p>\n<p><b>About Zilliz<\/b><\/p>\n<p>Zilliz is the company behind <u><a href=\"https:\/\/milvus.io\/?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Milvus<\/a><\/u>, the world&#8217;s most widely adopted open-source vector database. <u><a href=\"https:\/\/zilliz.com\/cloud?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Zilliz Cloud<\/a><\/u> brings that performance to production with a fully managed, cloud-native platform built for scalable, low-latency vector search and hybrid retrieval. It supports billion-scale workloads with sub-10ms latency, auto-scaling, and optimized indexes for GenAI use cases like semantic search and RAG.<\/p>\n<p>Zilliz is built to make AI not just possible\u2014but practical. With a focus on performance and cost-efficiency, it helps engineering teams move from prototype to production without overprovisioning or complex infrastructure. Over 10,000 organizations worldwide rely on Zilliz to build intelligent applications at scale.<\/p>\n<p>Headquartered in Redwood Shores, <span class=\"xn-location\">California<\/span>, Zilliz is backed by leading investors, including Aramco&#8217;s Prosperity 7 Ventures, Temasek&#8217;s Pavilion Capital, Hillhouse Capital, 5Y Capital, Yunqi Partners, Trustbridge Partners, and others. Learn more at\u00a0 <u><a href=\"https:\/\/zilliz.com\/?utm_source=vendor&amp;utm_medium=referral&amp;utm_campaign=seonews-Bilingual-semantic-highlighting-model\" target=\"_blank\" rel=\"nofollow\">Zilliz.com<\/a><\/u>.<\/p>\n<p>\u00a0<\/p>\n<div class=\"PRN_ImbeddedAssetReference\">\n<\/div>\n<p><!-- \/wp:html --><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"rop_custom_images_group":[],"rop_custom_messages_group":[],"rop_publish_now":"initial","rop_publish_now_accounts":[],"rop_publish_now_history":[],"rop_publish_now_status":"pending","footnotes":""},"categories":[1],"tags":[],"class_list":["post-47164","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/thaipropertynews.com\/feeds\/index.php?rest_route=\/wp\/v2\/posts\/47164","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/thaipropertynews.com\/feeds\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/thaipropertynews.com\/feeds\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/thaipropertynews.com\/feeds\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/thaipropertynews.com\/feeds\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=47164"}],"version-history":[{"count":0,"href":"https:\/\/thaipropertynews.com\/feeds\/index.php?rest_route=\/wp\/v2\/posts\/47164\/revisions"}],"wp:attachment":[{"href":"https:\/\/thaipropertynews.com\/feeds\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=47164"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/thaipropertynews.com\/feeds\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=47164"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/thaipropertynews.com\/feeds\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=47164"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}