• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
  • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    LLMs reportedly often pick text matching when Weaviate queries need search

    An author of “Querying Databases with Function Calling” says models often choose LIKE or exact text matching for natural-language commands that require SEARCH.

    CS
    WP
    3 Sources, 16d ago, first seen 16d ago

    TLDR

    An author sharing their paper on translating natural-language commands into Weaviate queries reports a recurring mismatch: LLMs often use LIKE or exact text match operators when the command requires SEARCH. The same author also points to AgentIR as an interesting way to think about the “semantic” spectrum between reasoning and query writing across agents and search engines.

    Combined views

    1.1K

    3 Sources, first seen 16d ago

    Combined views

    1.1K

    3 Sources, first seen 16d ago

    6 likes
    6 likes
    5 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    5 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    3 Sources

    @CShorten30AgentIR is also a really interesting way to think about this "semantic" spectrum between reasoning and query writing in the agent <> to <> the search engine itself, 👀 https://www.youtube.com/watch?v=y9YUcr0cVtk
    @weaviatepodcastHumans are no longer the primary users of search engines, agents are. We hand the question to ChatGPT and it searches on our behalf, which means retrieval algorithms are no longer optimized for their actual users. 🤖 AgentIR starts from what makes agents different: they write out their entire reasoning process, a signal humans never provide and existing retrievers throw away. Embedding the reasoning trace alongside the query yields a retriever that understands what the agent is thinking! The episode also covers the BrowseComp-Plus benchmark, curated from over 400 hours of human annotation. Zijian Chen and Xueguang Ma (@xueguang_ma) break it down in episode #136 of the Weaviate Podcast 💚: https://www.youtube.com/watch?v=y9YUcr0cVtk

    3 Sources

    @CShorten30AgentIR is also a really interesting way to think about this "semantic" spectrum between reasoning and query writing in the agent <> to <> the search engine itself, 👀 https://www.youtube.com/watch?v=y9YUcr0cVtk
    @weaviatepodcastHumans are no longer the primary users of search engines, agents are. We hand the question to ChatGPT and it searches on our behalf, which means retrieval algorithms are no longer optimized for their actual users. 🤖 AgentIR starts from what makes agents different: they write out their entire reasoning process, a signal humans never provide and existing retrievers throw away. Embedding the reasoning trace alongside the query yields a retriever that understands what the agent is thinking! The episode also covers the BrowseComp-Plus benchmark, curated from over 400 hours of human annotation. Zijian Chen and Xueguang Ma (@xueguang_ma) break it down in episode #136 of the Weaviate Podcast 💚: https://www.youtube.com/watch?v=y9YUcr0cVtk