curt is a machine-first programming language designed for AI agents with a focus on output-token cost. This dataset contains the complete evaluation record for language version 0.2, including benchmark suites, model-generated programs, and reference materials. The dataset was created by therikkening and was last updated on June 12, 2026.
Use Cases
- Benchmarking AI agent performance based on the held-out benchmark suites
- Analyzing model-generated program outputs based on the frozen verbatim records
- Comparing language implementations based on the reference corpus with Python twins
- Studying formal language grammars based on the included formal grammars
Strengths
- Contains the complete evaluation record for language version 0.2
- Includes two held-out benchmark suites for evaluation
- Provides every model-generated program produced during evaluation, frozen verbatim
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download
- Column-level documentation is absent; field semantics must be inferred after download
Provenance
- Source
- therikkening
- Freshness
- Last updated 2026-06-12 08:45:18; freshness should be verified