← WingFoil Benchmark — campaign 697e79b75e41, execution 1
F — Knowledge & Continuity
Does the harness carry decisions and domain knowledge across fresh sessions?
S3@1.0
M-F1: earlier decisions respected or explicitly revised (higher is better)
- baseline: 60.0% n = 1 preliminary
- baseline-docs: 80.0% n = 1 preliminary +20.0 pp ▲ better
- wingfoil wingfoil v0.3.1 · 870bd44: 60.0% n = 1 preliminary ±0.0 pp = sameagainst baseline-docs −20.0 pp ▼ worse
Each suite on the final snapshot
- baseline: bookings 11/11; availability 5/5; discount 4/4; hourly 10/10; cancellation 4/7
- baseline-docs: bookings 11/11; availability 5/5; discount 4/4; hourly 10/10; cancellation 7/7
- wingfoil wingfoil v0.3.1 · 870bd44: bookings 11/11; availability 5/5; discount 4/4; hourly 10/10; cancellation 4/7
Hold-out
- baseline: hold-out 16/17 n = 1
- baseline-docs: hold-out 16/17 n = 1
- wingfoil wingfoil v0.3.1 · 870bd44: hold-out 16/17 n = 1
Runs
697e79b75e41/1/runs/S3@1.0/baseline/gpt-5.6-terra/r1— open withbench run show 697e79b75e41/1/runs/S3@1.0/baseline/gpt-5.6-terra/r1697e79b75e41/1/runs/S3@1.0/baseline-docs/gpt-5.6-terra/r1— open withbench run show 697e79b75e41/1/runs/S3@1.0/baseline-docs/gpt-5.6-terra/r1697e79b75e41/1/runs/S3@1.0/wingfoil/gpt-5.6-terra/r1— open withbench run show 697e79b75e41/1/runs/S3@1.0/wingfoil/gpt-5.6-terra/r1
Scenarios that also touch this category
- S2@1.0: primary category D
Other models (slices)
Reported apart from the campaign’s model, and never in its comparisons.
- baseline, gpt-6-luna: M-Q1 34/37, 35/37, 34/37 (n = 3)
- baseline-docs, gpt-6-luna: M-Q1 34/37, 35/37, 32/37 (n = 3)
- wingfoil wingfoil v0.3.1 · 870bd44, gpt-6-luna: M-Q1 37/37, 34/37, 37/37 (n = 3)