Zeng et al.

Applying reinforcement learning with verifiable rewards to the GLM-4-Voice speech model raised free-form accuracy on the GSM8K math benchmark to 74.8%, a new state of the art for speech-native models,…

A developer found that when an LLM agent was given the same paired before/after dataset (n=25) five times and asked to determine if scores changed significantly, only one out of five runs checked the …