Add model mappings supported by Qwen's model card and Xiaomi's API transition notice. Keep API evaluation conditions explicit, preserve all eight frozen score rows, and record that GPT-6.1 Sol, Claude Opus 5 and Kimi K3.1 Pro currently lack matching official Hub repositories.
Configure native SLE leaderboard metrics and freeze published results
Declare Harbor as the evaluation framework and define five unambiguous 0–100 metrics for the 90-task capability scope. Add frozen eight-model values, exact Hub mappings for three API models, metric provenance and links from the dataset card. End-to-end and Hypothesis Discovery remain outside the composite; repository access is unchanged.
