Volcano's node-scoring pipeline combines several plugin stages into the score buckets used by allocation and…
Volcano's node-scoring pipeline combines several plugin stages into the score buckets used by allocation and…: a task in MiMo-V2.6-RL-oss: Agentic RL Environments (MiMo RL release). At present, a node whose per-node scorer fails can silently come back through a later batch score, invalid…
The task
At present, a node whose per-node scorer fails can silently come back through a later batch score, invalid floating-point values can become selectable buckets, all-negative rankings can produce no winner, and equal scores are resolved randomly. This makes plugin failures unsafe…
Part of XiaomiMiMo/MiMo-V2.6-RL-oss.