MindLink-32B and MindLink-72B available on Huggingface

Built on Qwen, these models incorporate our latest advances in post-training techniques. MindLink demonstrates strong performance across various common benchmarks and is widely applicable in diverse AI scenarios.

72B 32B

20 points · 2 comments · view on lemmy.world

2 Comments

humanspiral@lemmy.ca · 1 pts · 1y (1 reply)

reddit suggested "post training" means training on tests, rather than improving general programming outputs.

pepperfree@sh.itjust.works · 1 pts · 1y

Skywork downfall