Repair hierarchical ML evaluator aggregation and produce an offline integration artifact.
Repair hierarchical ML evaluator aggregation and produce an offline integration artifact.: a task in MiMo-V2.6-RL-oss: Agentic RL Environments (MiMo RL release). The evaluator in /app/vendor/lm-evaluation-harness has a regression in hierarchical metric aggregation. Diagnose the interaction between…
The task
The evaluator in /app/vendor/lm-evaluation-harness has a regression in hierarchical metric aggregation. Diagnose the interaction between weight_by_size, point estimates, and reported standard errors, and repair the existing production modules in lm_eval/api/metrics.py and…
Part of XiaomiMiMo/MiMo-V2.6-RL-oss.