mirror of
https://github.com/hanzoai/benchmarks.git
synced 2026-10-30 14:59:11 +00:00
Cardinality to 100,000 options, option-order and alias invariance, joint
programs, sensor states and the original harness rows, with the Jev spend
ledger. These are the artifacts data/runs.json names for the Kai paper's
capability tables; Kai's rows land with its trained checkpoint. Generated
cases and downloaded data stay ignored, as the directory's .gitignore says.
The b9235d3 message miscounted the Jev run: 9,499 cases, 9,498 answered, not
10,800.
Co-authored-by: Hanzo Dev <[email protected]>