EUNO.NEWS EUNO.NEWS
  • All (15818) +246
  • AI (2489) +20
  • DevOps (703) +14
  • Software (8152) +125
  • IT (4437) +85
  • Education (37) +2
  • Notice
  • All (15818) +246
    • AI (2489) +20
    • DevOps (703) +14
    • Software (8152) +125
    • IT (4437) +85
    • Education (37) +2
  • Notice
  • All (15818) +246
  • AI (2489) +20
  • DevOps (703) +14
  • Software (8152) +125
  • IT (4437) +85
  • Education (37) +2
  • Notice
Sources Tags Search
한국어 English 中文
  • 15小时前 · ai

    MiroMind的MiroThinker 1.5在30B模型上实现万亿参数性能——成本仅为原来的1/20

    加入越来越多的小型、强大推理模型行列的是来自 MiroMind 的 MiroThinker 1.5,只有 300 亿参数,相比于 h...

    #large language models #MiroThinker #MiroMind #parameter efficiency #cost reduction #AI performance
  • 4天前 · ai

    超越 Benchmaxxing:为何 AI 的未来在于推理时搜索

    文章 URL: https://adlrocha.substack.com/p/adlrocha-beyond-benchmaxxing-why 评论 URL: https://news.ycombinator.com/item?id=46486290 积分: 3 评论: 0

    #inference-time search #AI performance #model benchmarking #LLM optimization #machine learning inference
EUNO.NEWS
RSS GitHub © 2026