coopeval-static-eval

coopeval-static-eval

All 36 declared native panels are available as separately frozen evaluation leagues. Each submitted policy runs one complete selected panel per episode. Panel scores retain their native aggregation; this release defines no score across different panels. The fleet.json index records every panel, population, metric, and fixed reference roster. Qualification and coverage must be established separately for every selected panel.

CoopEval

Each seat controls a policy across all native game roles. Each episode runs the complete upstream tournament for one game and mechanism. Variants retain payoff matrices, rounds, discounting, and history settings. Scores are the upstream average policy payoffs; they are not paper-wide rankings. Replay retains provider responses, format-repair attempts, and native payoff and game records. Certification uses the upstream offline TestModel, explicitly labeled qualification/coopeval-control-v1. Submit pinned real players through the downstream suite for model evaluation. Source: https://github.com/Xiao215/CoopEval/tree/8559e858b5cc33e0a29ac592c0b21eab007feb57