Selection, Recombination, or a Fresh Solve? A Candidate-Free Control for Single-Pass Test-Time Aggregation

A new arXiv preprint (2608.18379v1) introduces a candidate-free control for single-pass test-time aggregation, finding that on AIME-2025 and HMMT-2025 with Qwen3-4B, conditioning on a candidate pool improves accuracy when multiple candidates are correct (Δ_cand(c2+) = +0.290) but lowers accuracy when every candidate is wrong (Δ_cand(c0) = -0.123), reversing the interpretation of all-wrong recovery at this scale. The authors note the evidence is limited to one model family, two benchmarks, and single-pass prompted aggregation.

arXiv:2608.18379v1 Announce Type: new Abstract: When every candidate is wrong, correct-candidate selection is unavailable, yet the aggregation call can still solve the problem afresh. A correct aggregate answer may therefore reflect recombination, fresh solving, or both. For efficient test-time reasoning, the relevant question is whether candidate context adds value beyond the additional generation pass. We introduce the missing candidate-free control under the same maximum output-token allowance and stratify by the number of correct candidates. Across AIME-2025 and HMMT-2025 with Qwen3-4B, candidate conditioning improves accuracy when multiple candidates are correct $\Delta {\mathrm{cand}}$ c2+ = +0.290 , lowers accuracy when every candidate is wrong $\Delta {\mathrm{cand}}$ c0 = -0.123 , and remains unresolved in the one-correct regime. The c2+ and c0 conclusions survive a conservative correction for the adaptive two-benchmark procedure. Under this counterfactual, the interpretation of all-wrong recovery reverses at this scale: conditioning on an all-wrong candidate pool lowers accuracy relative to a fresh solve. Original-format matching and placebo results characterize the failures descriptively but leave their mechanism unresolved. Within a separate structured intervention, explicit answer fields causally steer outputs toward their values; masking yields no measurable accuracy improvement, and equivalence with the original format was not established. The evidence is limited to one Qwen3-4B family, two mathematics benchmarks, first-answer-truncated candidate fragments, and single-pass prompted aggregation.