Web Machine Learning WG F2F – 10 November 2025 - meeting minutes

Hi All,

Thanks for joining our F2F!

Minutes:
https://www.w3.org/2025/11/09-webmachinelearning-minutes.html

Summary:

  1.  Welcome<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#b0a5>
     *   Intros<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#6e8c>
     *   Agenda bashing<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#6052>
     *   Charter orientation<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#1233>
  2.  Spec orientation<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#e89f>
  3.  WebNN Small Language Model (SLM) Performance Optimization Case Study<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#3e1a>
  4.  New features<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#b352>
     *   WebNN Small Language Model (SLM) Performance Optimization Case Study<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#da7d>
     *   Core operator set<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#23a8>
     *   Support flexible input sizes<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#5b68>
     *   API to Separate Graph Building from Weight Loading to Reduce Peak Memory Usage<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#2978>
     *   Device selection, state of the union<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#11ce>
     *   Device selection criteria for usecase-driven scenarios<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#93d2>
     *   CPU fallback hint<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#fae4>
     *   Get device selection information after graph compilation<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#d1f8>
     *   MLOpSupportLimits<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#040b>
  5.  Customer feedback & collaborations<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#f588>
     *   RTC-style workloads with response time requirements<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#43b5>
  6.  Implementation plans and trials<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#5ca9>
     *   Browser vendors' trials<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#c193>
  7.  Horizontals<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#36cc>
     *   Ethics<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#0949>
     *   Sustainability<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#47e8>
     *   Privacy and Security<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#aa23>
  8.  Wrap up<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#6962>
  9.  Summary of resolutions<https://www.w3.org/2025/11/09-webmachinelearning-minutes.html#ResolutionSummary>

Thanks,

-Anssi (WebML WG/CG chair)