Computed Retrieval, Decided Verdicts: A Parameter-Free Retrieval-to-Verification Chain over a Decidable Program Fragment

Dennis Marvin Lobb

doi:10.5281/zenodo.21984818

August 17, 2026Open Access

Computed Retrieval, Decided Verdicts: A Parameter-Free Retrieval-to-Verification Chain over a Decidable Program Fragment

View Full Paper

Ask AI

Bookmark

Share

Key Points

To introduce and evaluate the LNFdex engine, a parameter-free retrieval-to-verification chain that pairs hash-derived binary hypervector retrieval with exhaustive verification over a decidable program fragment.
Designed a computed retrieval mechanism using hash-derived 256-bit binary hypervectors and Hamming ranking without learned parameters or floating-point operations.
Implemented exhaustive verification with sound and complete abstention across a 54-point decidable program fragment.
Benchmarked execution performance on a Raspberry Pi 5 against the PLANWERK_v0_1_24_PI5 artifact, comparing performance with three small language models and executing a 10-language render/parse round-trip across 540 point checks.
Reduced verification overhead from 54 checks to 1 exact check at 0 model tokens.
Achieved a 20/20 success rate in 78 ms, compared to 0/20 for three small language models evaluated on the same test harness.
Demonstrated proven correctness across all 540 point checks in the 10-language render/parse round-trip evaluation.

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

Benchmarking analysis demonstrates deterministic verification in an embedded runtime environment, highlighting parameter-free hypervector retrieval superiority over language models.

Abstract

Technical report of the LNFdex(TM) engine: a retrieval-to-verification chain in which retrieval is computed (hash-derived 256-bit binary hypervectors, Hamming ranking, no learned parameters, no floats) and the verdict is decided (exhaustive verification over a 54-point decidable program fragment with sound and complete abstention). Measured on Raspberry Pi 5 against the frozen artifact PLANWERK_v0_1_24_PI5: 1 exact check instead of 54 at 0 model tokens; 20/20 in 78 ms vs. 0/20 for three small language models on the same harness; 10-language render/parse round-trip proven over all 540 point checks. Companion to the P1 paper line.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Dennis Marvin Lobb (2026) studied this question.

synapsesocial.com/papers/6a85638803308d306e2d6cf1 https://doi.org/10.5281/zenodo.21984818

View Full Paper

Ask AI

Bookmark

Share

Computed Retrieval, Decided Verdicts: A Parameter-Free Retrieval-to-Verification Chain over a Decidable Program Fragment

Key Points

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Computed Retrieval, Decided Verdicts: A Parameter-Free Retrieval-to-Verification Chain over a Decidable Program Fragment

Key Points

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading