Xiaomi released MiMo-V2.6 weights under MIT and, unusually, disclosed the RL bill: $3.47 million for under six days of training.
A Tsinghua team introduced SEED — self-evolving on-policy distillation with hindsight supervision. On ALFWorld it improves success by 45.9 points over GRPO and matches it using 60% of the data.
