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

    NeoMME-Retriever-260M tops evaluated sub-800M models on ViDoRe v3, a user reports

    Using late-interaction embeddings, the 260M-parameter model scored 0.523 nDCG@10, while the 800M model scored 0.556, according to the user.

    OK
    TW
    TO
    5 Sources, ,

    TLDR

    A user reports that NeoMME-Retriever-260M reached 0.523 nDCG@10 on ViDoRe v3 using late-interaction embeddings—the highest score among evaluated models below 800 million parameters. nDCG@10 measures ranking quality across the top 10 results. The user reports a score of 0.556 for the 800M model.

    Combined views

    10.6K

    5 Sources, first seen 27d ago

    Combined views

    10.6K

    5 Sources, first seen 27d ago

    83 likes
    27d ago
    first seen 27d ago
    83 likes
    5 comments
    36 saves
    21 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    5 comments
    36 saves
    21 reposts
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    5 Sources

    @tonywu_71On ViDoRe v3 and using the late-interaction embeddings, NeoMME-Retriever-260M reaches 0.523 nDCG@10, the highest score among evaluated models below 800M parameters. The 800M model reaches 0.556. Both are on the benchmark's model-size Pareto frontier. (7/N)
    @tomaarsen🔎 @hcompany_ai just released NeoMME: 260M & 800M multilingual encoders for text and images. You can load them in Sentence Transformers to search document pages with MultiVectorEncoder, charts & tables included, without an OCR step. Thread 🧵
    @lateinteractionRT @tomaarsen: 🔎 @hcompany_ai just released NeoMME: 260M & 800M multilingual encoders for text and images. You can load them in Sentence T…

    5 Sources

    @tonywu_71On ViDoRe v3 and using the late-interaction embeddings, NeoMME-Retriever-260M reaches 0.523 nDCG@10, the highest score among evaluated models below 800M parameters. The 800M model reaches 0.556. Both are on the benchmark's model-size Pareto frontier. (7/N)
    @tomaarsen🔎 @hcompany_ai just released NeoMME: 260M & 800M multilingual encoders for text and images. You can load them in Sentence Transformers to search document pages with MultiVectorEncoder, charts & tables included, without an OCR step. Thread 🧵
    @lateinteractionRT @tomaarsen: 🔎 @hcompany_ai just released NeoMME: 260M & 800M multilingual encoders for text and images. You can load them in Sentence T…