events
Paper Reading: Matryoshka Attribution
Lettura online di un articolo su Matryoshka attribution, un metodo per attribuire gli output dei modelli linguistici alle rappresentazioni e ai pesi.

In inglese
- Starts
- 2026-10-13 00:30 UTC
- Organizer
- Silicon Valley Generative AI ~ The AI Collective Network
- Location
- Online event
An online paper reading session hosted by Silicon Valley Generative AI, part of the AI Collective Network, in collaboration with Boulder Data Science. Logan presents "Matryoshka attribution: Learning to attribute language model outputs to representations and weights," a paper on LLM attribution and interpretability using learned matryoshka masks. The group meets every second week to work through recent machine learning papers. The session is aimed at a technical audience comfortable with language model internals; a paper link and code repository are shared beforehand.
Event website