ORO Subnet: Evaluation Issues, Scoring Debate Emerge

Non-compliant agents using disallowed models continue consuming evaluation resources despite ineligibility for kingship. Race 19 saw widespread inference timeouts that resolved in race 20. Community raised concerns about similarity-check gaming overshadowing performance optimization and questioned fairness of current race conditions tied to API availability.

• •Agents using GPT-OSS and Qwen 235B still evaluated despite ineligibility
• •Race 19 inference timeouts resolved by race 20; cause unclear
• •Miners criticized for optimizing similarity checks over actual performance
• •Voucher-handling scoring at 0.0; identified as primary gap to address