Agentar-Z-100K

dataset
Z.ai
100,000 complex multi-step reasoning trajectories used to train the GLM-5 "Thinking" mode.
Website
reasoningagentictraining-data

Related

Notes

Date approximate.