Six Adaptive Queries for Fixed-Point Detection in Two-Image Maps
Year: 2026
Type: preprint
Abstract: We study adaptive evaluation-oracle detection of a fixed point in finite self-maps (f:X\to X) whose image contains at most two elements, allowing at most one adversarially substituted oracle response over the entire query sequence. We present a frozen 101-rule equality-history decision strategy, together with a reproducible finite verification certificate for universe sizes (3\leq n\leq14). The computational evidence includes explicit adversarial replay for (n=3,4,5), independent symbolic verification procedures, negative controls, and an exact three-state minimax computation establishing a six-query lower bound in the finite base model. We formulate mathematical completion, canonical-history, and support-reduction arguments connecting the finite verification to the proposed universal six-query result. In particular, the support-reduction argument shows how a hypothetical failing execution involving at most six queries can be represented using at most fourteen states while preserving the fixed-point property and the adversarial error budget. The work is presented as a non-peer-reviewed, computer-assisted mathematical preprint. Although the computational certificates have passed the reported verification procedures, the mathematical reductions and verifier implementation remain open to independent specialist scrutiny. A targeted literature review situates the study in relation to Rényi–Ulam searching games, adaptive query complexity, and black-box fixed-point problems. The accompanying research package contains the manuscript, Python source code, an independent verification program, reproduction instructions, licensing information, and SHA-256 checksums. This release contains a non-peer-reviewed mathematical preprint and its reproducibility package. The computational certificate verifies a frozen 101-rule adaptive strategy for finite universe sizes 3–14 and includes explicit adversarial replay and an independent minimax computation. The universal mathematical conclusion depends on the correctness of the accompanying completion, canonical-history, and support-reduction arguments. Independent mathematical peer review has not yet been completed. AI tools assisted with mathematical exploration, programming, manuscript preparation, and literature discovery. The author takes responsibility for the submitted work. The manuscript is licensed under CC BY 4.0, while the accompanying Python source code is licensed separately under the MIT License.
Author: Vajc, Ivo
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236507319
