LESSWRONG
LW

397

MATS Program

Edited by Ryan Kidd, Multicore, et al. last updated 30th Dec 2024

ML Alignment & Theory Scholars (MATS) Program is an educational seminar and independent research program that aims to provide talented scholars with talks, workshops, and research mentorship in the field of AI alignment, and connect them with the Berkeley AI safety research community.

2

2

Posts tagged MATS Program

87Trends in Economic Inputs to AI

Jeffrey Heninger

1mo

6

46Apply to MATS 9.0!

1mo

0

22MATS 8.0 Research Projects

Jonathan Michala, DanielFilan, Ryan Kidd

1mo

0

14Evaluating Prediction in Acausal Mixed-Motive Settings

1mo

0

57Discovering Backdoor Triggers

andrq, Tim Hua, Sam Marks, Arthur Conmy, Neel Nanda

2mo

4

53Towards data-centric interpretability with sparse autoencoders

Nick Jiang, lilysun004, lewis smith, Neel Nanda

2mo

2

132Training a Reward Hacker Despite Perfect Labels

ariana_azarbal, vgillioz, TurnTrout

2mo

45

62Statistical suggestions for mech interp research and beyond

2mo

4

59Concept Poisoning: Probing LLMs without probes

Jan Betley, jorio, dylan_f, Owain_Evans

2mo

5

16Exploration hacking: can reasoning models subvert RL?

Damon Falck, Joschka Braun, Eyon Jang

2mo

4

196Optimizing The Final Output Can Obfuscate CoT (Research Note)

lukemarks, jacob_drori, cloud, TurnTrout

2mo

22

55The many paths to permanent disempowerment even with shutdownable AIs (MATS project summary for feedback)

2mo

6

39Building Black-box Scheming Monitors

james__p, richbc, Simon Storf, Marius Hobbhahn

2mo

18

54Unfaithful chain-of-thought as nudged reasoning

Paul Bogdan, Uzay Macar, Arthur Conmy, Neel Nanda

3mo

3

51Do LLMs know what they're capable of? Why this matters for AI safety, and initial findings

Casey Barkan, Sid Black, Oliver Sourbut

3mo

5

Load More (15/248)

Add Posts