Skip to content
Sign in
Artwork for Models & Agents
Models & Agents · September 28 · 10 min

Ep 187: Sparse neuron sets in frozen BERT enable efficient AI-text detection across generators…

Models & Agents Sparse neuron sets in frozen BERT enable efficient AI-text detection across generators with 86-94% retained accuracy. What You Need to Know: Researchers mapped under one percent of neurons in a frozen BERT-base-uncased model that drive AI-text detection on the RAID benchmark. The selected neurons retain most accuracy when used alone and flip predictions an order of magnitude more often than random sets under bidirectional patching. ... Sources: arxiv.org AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. 🎬 Watch on YouTube: https://www.youtube.com/watch?v=ix91NIyaRdc 📝 Full show notes, transcript & sources: read the episode page 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.

0:00 · Introduction-10:54
Play in
Transcript file

transcript

No transcript — this publisher did not publish one.

show notes

Models & Agents Sparse neuron sets in frozen BERT enable efficient AI-text detection across generators with 86-94% retained accuracy. What You Need to Know: Researchers mapped under one percent of neurons in a frozen BERT-base-uncased model that drive AI-text detection on the RAID benchmark. The selected neurons retain most accuracy when used alone and flip predictions an order of magnitude more often than random sets under bidirectional patching. ... Sources: arxiv.org AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis for audio production. 🎬 Watch on YouTube: https://www.youtube.com/watch?v=ix91NIyaRdc 📝 Full show notes, transcript & sources: read the episode page 🌐 Part of the Nerra Network — explore every show at nerranetwork.com.
links3

chapters

7 chapters

more episodes

All episodes