Datasets:
Tasks:
Text Generation
Languages:
English
Size:
n<1K
Tags:
structured-editing
structured-edit-interface-study
language-model-interfaces
verified-code-edits
fail-closed-evaluation
software-engineering-agents
License:
File size: 743 Bytes
4062a7b | 1 2 3 4 5 6 7 8 | # Methods
The release dataset is copied from the sanitized public package and augmented only with release-candidate metadata, cards, audit files, and manifests. It is provider-free and does not construct network transport.
For multimodel records, strict native and one-outer-fence-normalized policies are reported on identical response bytes. DeepSeek V4 Pro cells are infrastructure-unassessable privacy/data-policy route failures with empty final answers, not coding-performance failures.
Low-affinity calibration tasks are newly authored for calibration where supported by provenance. High-affinity calibration tasks are adapted or reconstructed from upstream issue/task patterns; upstream source bytes and evaluator bytes are withheld.
|