Weaviate 1.39 上线 MMR 多样性选择，平衡相关性与覆盖面
Relevance is not the same thing as coverage. For the query "japan", standard vector search can retu...
做 RAG 检索的可以看看：Weaviate 1.39 的 MMR 参数一个 balance 就能让搜索结果不扎堆，且不用重建索引。
Weaviate 1.39 将 MMR 多样性选择正式开放到所有 near_* 搜索和 hybrid 搜索。以查询 "japan" 为例，普通向量搜索可能返回五张彼此近似重复的图片，而 MMR 在查询时逐个贪心选取，同时兼顾与查询的相似度和与已选结果的相似度。balance=1.0 时排名保持纯相关性，balance=0.5 时首页可以覆盖富士山、火车、猴子、拉面、鸟居等不同内容。整个过程在查询时运行，已有集合无需改 schema 或重建索引。需要注意的是要显式传入 balance 参数，其默认值 0.0 偏向纯多样性，可能偏离查询意图。
Relevance is not the same thing as coverage. For the query "japan", standard vector search can retu...
Relevance is not the same thing as coverage. For the query "japan", standard vector search can return five highly relevant images that are also near-duplicates of each other. Hybrid search can hit the same problem when keyword and vector results converge on one dense cluster. MMR diversity selection reduces that repetition at query time. It selects results greedily, one at a time, balancing: • similarity to the query • similarity to results already selected With 𝗯𝗮𝗹𝗮𝗻𝗰𝗲=𝟭.𝟬, ranking stays focused on pure relevance. At `balance=0.5`, that first page can spread across Mount Fuji, a train, a monkey, ramen, and torii gates. MMR is now generally available in Weaviate 1.39 across every 𝗻𝗲𝗮𝗿_* search and hybrid search. It runs entirely at query time, so existing collections need no schema changes or reindexing. One important caveat: pass 𝗯𝗮𝗹𝗮𝗻𝗰𝗲 explicitly. Its default is 𝟬.𝟬, which prioritizes pure diversity and can wander surprisingly far from th weaviate.io/blog/weaviate-… ttps://t.co/Azo67V9iUy 💬 0 🔄 0 ❤️ 0 👀 60 ⚡