FUTURE-TS docs
Start here:
- Benchmark card: what FUTURE-TS v0.1.0 measures, the current benchmark surface, current empirical results, and scope limits.
- Submission guide: how to submit a new model entry.
- Submission spec: the JSON contract for submissions.
- Sealed runner: the execution and integrity contract.
- Validity envelope: what this release supports and what should not be overclaimed.
Design references:
- Benchmark design: task tiers, task cards, metrics, and aggregation philosophy.
- Live tier: future-only evaluation semantics.
- Prequential cutoffs: rolling live-tier scoring.
- Design journal: implementation history and rationale.