VAM! AI Reading Group: How Logical are LLMs?
🤖 AI Reading Group Meetup: Details 🤖
Next Reading is on "LogicBench: Towards Systematic Evaluation of Logical Reasoning Ability of Large Language Models"
Paper: https://arxiv.org/pdf/2404.15522
A cool paper that introduces LogicBench, a dataset designed to test how well AI models handle different types of logical reasoning such as complex logic, negations, and understanding context, highlighting areas for improvement.
🚀 Key Takeaways:
LLM Compiler, built on Code Llama, is trained on massive datasets of LLVM-IR & assembly code.
Can help automate code optimization, making software faster and more efficient.
Reduces the need for manual tuning, saving time and effort for developers.
This meetup will be held online only.
How to Prepare:
Please try to read the paper before the meetup. Don’t worry if you don’t understand everything—we’ll start with a short presentation to explain the key ideas and then have a group discussion.
You can use NotebookLM to have a quick overview of the paper: https://notebooklm.google/
How to Join:
Join via the provided Zoom link
Schedule:
6:00 PM Sharp: Meetup starts on time.
20-Minute Presentation: Summary of the paper and key points.
25-Minute Discussion: Group Q&A and sharing of ideas.
About the AI Reading Group:
We meet every week to discuss interesting AI topics and papers.
For past and proposed papers, check this spreadsheet: AI Reading Group Papers.
Anyone can volunteer to present the next paper. You can get ideas from Hugging Face Daily Papers.
This is a great opportunity to learn about AI and connect with others. Hope to see you there!