Oro Reasoning Judge Updated, Validators Fixed

Oro shipped a reasoning quality scoring refinement today with a clearer 5-bin structure and refreshed judge model pool (GLM-5.1, GLM-5, gemma-4-31B-turbo, Kimi-K2.5) to improve score reproducibility and judge availability. Validators experienced downtime mid-day but were restored; allowlisted models will update after today's race, and miners will soon gain the ability to pin specific agent versions for races. Multiple miners reported local test environment failures and permission errors after pulling new docker/git repos.

• •Reasoning judge uses 5-bin scoring (0.1–0.9) tied to specific evidence criteria
• •Judge model pool refreshed for tighter cross-model agreement and availability
• •Allowlisted models updating after today's race; agent pinning feature coming soon
• •Local sandbox test failures reported; validator downtime fixed same day