The standard in-sample metric for test-time allocation in neural combinatorial optimization manufactures 2 to 3 percent phantom gains, but a split-sample correction removes them and preserves real gains under distribution shift.
Inference on winners,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Sampling Luck Masquerades as Allocation Gain: Auditing Test-Time Budget Allocation for Neural Combinatorial Optimization
The standard in-sample metric for test-time allocation in neural combinatorial optimization manufactures 2 to 3 percent phantom gains, but a split-sample correction removes them and preserves real gains under distribution shift.