Samuel Albanie @samuelalbanie.bsky.social · 28/12/2024Video summary of deliberative alignment youtu.be/1efVS4DeEOs Links: - Paper: arxiv.org/abs/2412.16339 - Blog: openai.com/index/delibe...youtu.beDeliberative AlignmentYouTube video by Samuel Albanie 021
Samuel Albanie @samuelalbanie.bsky.social · 27/12/2024Video summary of recent work on alignment faking www.youtube.com/watch?v=_1bz...youtube.comAlignment Faking in Large Language ModelsYouTube video by Samuel Albanie 031
Samuel Albanie @samuelalbanie.bsky.social · 15/12/2024Had a great time at NeurIPS Thank you to everyone I got to talk to, especially at the poster sessions And thanks to the organizers for picking a beautiful location (the video is from a nearby hike with Vikrant) www.youtube.com/watch?v=MBGI...youtube.comStawamus Chief Trail, British Columbia (2024)YouTube video by Samuel Albanie's Miscellany 010
Samuel Albanie @samuelalbanie.bsky.social · 15/12/2024Clearly, I took the #runconference seriously. 020
Samuel Albanie @samuelalbanie.bsky.social · 14/12/2024How does data scale influence performance NeurIPS 2024 poster presentation By @vishaalurao.bsky.social youtu.be/YNZ23YPasXoyoutu.beNeurIPS 2024 Poster - No "Zero-Shot" Without Exponential DataYouTube video by Samuel Albanie 010
Samuel Albanie @samuelalbanie.bsky.social · 09/12/2024The GRAB benchmark Work with Jonathan Roberts and Kai Han youtu.be/XW3YdNATjIUyoutu.beStill a long way to go for Computer Vision? The GRAB BenchmarkYouTube video by Samuel Albanie 010
Samuel Albanie @samuelalbanie.bsky.social · 02/12/2024Will be at NeurIPS next week. DM if you're interested in meeting up for a chat (or a jog). 010
Reposted by Samuel AlbanieVishaal Udandarao @vishaalurao.bsky.social · 02/12/2024🚀New Paper: Active Data Curation Effectively Distills Multimodal Models arxiv.org/abs/2411.18674 Smol models are all the rage these days & knowledge distillation (KD) is key for model compression! We show how data curation can effectively distill to yield SoTA FLOP-efficient {C/Sig}LIPs!! 🧵👇 1236
Reposted by Samuel AlbanieArthur Douillard @douillard.bsky.social · 01/12/2024PrimeIntellect have released their tech report on INTELLECT-1: t.co/8hnoTILaL3 The first open-source world-wide training of a 10B model. The underlying ML distributed algo is DiLoCo (arxiv.org/abs/2311.08105) but they also built tons of engineering on top of it to make it scalable. 0121
Samuel Albanie @samuelalbanie.bsky.social · 23/11/2024This is a nice benchmark for AI R&D LLMs are closing the gap to humans Details: metr.org/AI_R_D_Evalu... 010
