Why can't I see the model generate multiple calls for conduct research simultaneously？ #226
Description
Hi experts.
I'm currently working on some optimizations for concurrent agent tasks. I noticed that in the code of open deep research, the supervisor generates multiple conduct_research_calls and then concurrently executes multiple researches. I think this is a very good design. However, in the actual operation of the agent, I have never seen the model generate more than one tool call here at the same time. I've tried many cases and it's still the same. The model I use is gpt-oss-120B. May I ask if the open deep research team has discovered that any case generates multiple tool calls simultaneously?
Looking forward to your reply.