Autoresearch Log 001

​Autoresearch Log 001: Automating LLM Post-Training

Paper club and discussions for people working on autoresearch and AI scientists: AI systems that run research loops on their own. They propose ideas, run experiments and check the results.

​For scientists, engineers and founders who are building these systems, or who have a problem they'd like to point one at. Talks are open discussions, with interruptions, questions and push back at any point.

​Program:

• ​Main talk: Maksym Andriushchenko from the ELLIS Institute Tübingen, one of the authors of PostTrainBench.com. A 15-minute talk with open discussion

• ​Paper club: ~3 papers, each with a 5-minute summary and 10 minutes of discussion

• ​Build Logs: quick demos of experiments people have built and run

• ​Networking, open end

​Papers

​The papers come from you. If there's a paper you'd like to present and discuss, tell us when you sign up. It doesn't have to be your own. (Topic does not have to align with the main talk)

​Demos

​Built something with autoresearch, or tried and hit a wall? Show it in a few minutes. A rough setup, a surprising result or an experiment that failed in an interesting way are all good demos. Describe yours briefly when you sign up.

This first edition of the Autoresearch Log is co-hosted by Inventionistas and the ETH AI Club