NEAR AI Paper Club -s1: Simple Test-time Scaling

​Join us for NEAR AI's paper club as we explore "s1: Simple test-time scaling", an innovative approach to improving language model performance through test-time compute optimization.

​Event Details:

• ​Date: Tuesday, March 4th, 2025

• ​Time: 6:30pm - 8:00pm PT

• ​Format: Discussion-based led by a moderator

• ​Location: NEAR AI Office

​Key Topics We'll Discuss:

• ​A minimal approach to achieving test-time scaling and strong reasoning performance

• ​The effectiveness of supervised fine-tuning with just 1,000 carefully curated samples

• ​Novel "budget forcing" technique for controlling test-time compute

• ​Experimental results showing performance comparable to OpenAI's o1-preview

• ​Implications for efficient model training and inference

​Before Joining:

• ​Please review the paper: https://arxiv.org/abs/2501.19393

• ​Come prepared with questions and insights to share

• ​Be ready to engage in constructive technical discussion

​Who Should Attend:

• ​AI researchers and practitioners

• ​Those interested in language model optimization

• ​Anyone curious about efficient ML training approaches

​About the Format:

• ​Interactive discussion environment

• ​Focus on technical details and practical implications

• ​Opportunity for Q&A and knowledge sharing