2024-10-28
- CognitionTO paper reading
- Event: https://lu.ma/rewscli9
- Paper: GSM-symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models
- Bathrooms are out the door, straight to the right
- notes
- Models have worse performance on GSM-Symbolic, even though it's basically just a draw of GSM8K
- data contamination?
- Models have worse performance on GSM-Symbolic, even though it's basically just a draw of GSM8K