3 ms·
My guess is that Gemini team didn't focus on the large-scale RL training for the agentic workload. And they are trying to catch up with 3.1.
by hintymad 8mo ago
My guess is that Gemini team didn't focus on the large-scale RL training for the agentic workload. And they are trying to catch up with 3.1.