A new model has been released to enhance the mathematical and reasoning capabilities of LLMs, specifically targeting the improvement of GSM8K scores. The model achieves a nearly 13% increase in GSM8K score and an overall improvement of about 1% on average.

3m read timeFrom blog.abacus.ai
Post cover image
Table of contents
Path to PerformanceSharpening LLM ReasoningInterleaving DPO and SFTKey Observations and The Road Ahead

Sort: