SRMs & Few-Step Discrete Flows: Connections between Algebraic Determinism and Probabilistic Generation

1 SRMs & Few-Step Discrete Flows: Connections between Algebraic Determinism and Probabilistic Generation Author: Carlos Roberto França 1[0000-0002-6852-7103] 1Federal University of Fronteira Sul – UFFS/Campus Chapecó-Santa Catarina – Brazil [email protected] Abstract This paper proposes a unifying interpretation between two fields that, although born in different times and contexts, move toward the same fundamental principle: few-step computational optimization. In 2017, the Heru Search Method (HSM), based on Infinite Series with Multiple Ratios (SRMs), was published in the American Journal of Computational Mathematics, presenting a deterministic method capable of replacing binary trees and breaking the century-old O (log n) paradigm. In 2025, Few-Step Discrete Flow-Matching (FS-DFM) emerged, developed by researchers at Apple Inc. and Ohio State University, introducing, in the domain of probabilistic generation, the same logic of discrete jumps with minimum procedural entropy. The objective of this article is to demonstrate that such systems — the deterministic (SRMs-HSM) and the probabilistic (FS-DFM) — are complementary expressions of the same universal law of efficiency: the Principle of Discrete Transition in Few Steps. The convergence proposed here does not aim to superimpose models, but to reveal the dual nature of the same mathematical-computational phenomenon: the transformation of the traversal into a jump. Keywords: SRMs. Heru Search Method.FS-DFM 1. Introduction Since the second half of the 1990s, Infinite Series with Multiple Ratios (SRMs) have been studied as an extension of arithmetic and geometric progressions and all of Newtonian kinematics. Their power to represent multiple relationships between elements of a series allowed, in 2017, the formulation of the Heru Search Method (HSM), an indexing and search model based on the concepts of the author's formulas called Infinite Series with Multiple Ratios (SRMs), capable of locating data without traversing binary paths. At that time, the concept of deterministic discrete transition was quietly inaugurated, in which data is not found by traversal, but by an algebraic leap of compound ratio. Eight years later, the global landscape witnessed the emergence of FS-DFM (Few-Step Discrete Flow-Matching), a model focused on text generation and probabilistic learning that also proposes replacing linear traversal (token by token) with discrete jumps in a few steps. Although in different domains, both systems, SRMs and FS-DFM, point to the same theoretical vector: reducing the path while maintaining the integrity of the information. 2 One does this through a deterministic algebraic approach, the other through a stochastic flow approach. The result is the birth of an unprecedented symmetry between what we could call FewStep Generation and Search. Figure 1 - SRMs & FS_DFM: Unified Few-Step Transation Framework Source: Author using GPT 5 OpenAI (2025) 2. Mathematical and Structural Foundations of Convergence The core of the FS-DFM proposal is based on mapping discrete distributions through reduced probability trajectories, flows that leave a "noise" state and reach the "given" state in a few steps. In SRMs, the movement is reversed, but analogous: one starts from a structured data domain, whose position is obtained by multiple ratio equations r₁, r₂, r₃…, and the transition between periods (P₁ → P₂ → P₃) is calculated deterministically, also in a few steps. 3 Both models, therefore, share: 1. Controlled flow discretization: in FS-DFM, via flow-matching scalar update; in HSM/SRMs, via multiple cumulative ratios. 2. Reducing operational entropy: replacing long paths with mathematically predictable jumps. 3. Unifying efficiency and integrity: neither sacrifices consistency for speed; both redefine it through structure. When analyzed from the perspective of computational physics or informational biology, SRMs act as a deterministic organizing field, while FS-DFM flows represent the statistical behavior of the same field at a probabilistic level. In short, they are two sides of the same equation. 3. Complementarity and the Few-Step Universe The Heru Search Method (HSM) [1] uses the foundations of Infinite Series with Multiple Ratios (SRMs) to reorganize and allocate data in a dynamic and structurally differentiated manner. This mathematical reprocessing step creates a logical mesh capable of distributing and indexing any type of numeric, textual, symbolic, or hybrid information, in periods and subperiods that follow the multiple ratios of the SRMs system. Thus, data is allocated in prepared structures, where each element knows its own "analytical address," eliminating the need for sequential traversal of binary trees. The result is a deterministic access organization, in which the number of steps is drastically reduced compared to the O (log n) paradigm, achieving efficiency equivalent to or greater than the O(1) of Hash Tables, but without collisions, losses, and without dependence on hash functions. This is one of the greatest points of convergence between the SRMs/HSM and FS-DFM frameworks. While FS-DFM proposes the reconstruction of probabilistic trajectories in a few steps for the purpose of generating information, the Heru Search Method operates on the deterministic basis of data allocation and reorganization in a few steps. Both share the same principle of computational trajectory compression, replacing long and redundant paths with discrete and predictable jumps. The great advantage of SRMs is their universal application: they can reorganize and optimize any type of data, not just text or linguistic tokens. This algebraic generalization makes the SRMs-HSM model not only compatible with, but also complementary and empowering to, FS-DFM, paving the way for a unified framework in which probabilistic generation and deterministic search coexist under the same few-step transition law. 4 4. Potential Applications and Technological Impact The convergence between FS-DFM and the Heru Search Method (HSM) opens up unprecedented horizons for the optimization of computational processes, combining the adaptive probabilistic power of flow generation with the deterministic precision of SRM structures. The field of application of this combination is vast: from the reconstruction of contexts in generative models to the structuring of self-organizing databases, collision-free indexing systems, algebraic data compression, bioinformatics, cryptography, and hybrid quantum computing. While Few-Step Discrete Flow-Matching seeks to reduce the number of iterations required to generate or reconstruct discrete data sequences, the Heru Search Method has been operating since 2017 [1] under the same principle, but focused on the organization and instantaneous retrieval of previously structured information. The difference is that, in SRMs, the path is not learned, but precisely calculated. The jump between states is a direct result of a multi-rational analytical function, not a statistical distribution. This characteristic gives the HSM superior structural stability, especially when compared to traditional indexing models such as Hash Tables. As detailed in Paper 7 [2], the HSM achieves the same O (1) access performance as Hash Tables, without incurring collisions, without relying on hash functions, and maintaining the natural ordering of the data. In practical terms, this means that the SRM-HSM achieves what Hash Tables promise, but without the hidden costs of re-hashization, collisions, or loss of predictability. Thus, Few-Step Discrete Flow-Matching and the Heru Search Method can be seen as two complementary implementations of the same universal computational efficiency law: • The FS-DFM represents the probabilistic jump applied to information generation, • And the HSM expresses the deterministic jump applied to data structuring and retrieval. The integration between both heralds a new era of unified optimization, in which the boundary between storage and generation dissolves, giving rise to continuous flow computing, where data and ideas inhabit the same transition field. 5. Conclusion — The Synthesis between Generation and Search as a Universal Computational Law Throughout this work, we have demonstrated that the computational phenomenon known as Few-Step Discrete Flow is not a recent coincidence, but rather the confirmation of a logic already manifested in Infinite Series with Multiple Ratios (SRMs) since 1996 [3] and in the Heru Search Method (HSM). What FS-DFM achieves through probabilistic 5 flows, SRMs already achieved through deterministic multiple-ratio transitions—two distinct paths that lead to the same mountain: the reduction of the path toward information. This synthesis reveals something greater than a theoretical coincidence. It points to the emergence of a Universal Computational Law, where the generation (output) and search (retrieval) processes begin to obey the same efficiency metric: the few-step. The act of generating a token and the act of locating a piece of data become symmetrical expressions of a single operation: the optimized discrete transition. In this context, FS-DFM and HSM do not compete; they complement each other. The former, born from statistics and diffusion, learns to reduce the number of steps by observing the flow. The latter, born from pure mathematics, reduces the number of steps by formulating the flow. One learns the path; the other creates it. Both prove that the future of computing will not be binary or continuous, but algebraic-dynamic. SRMs, when applied to the data space, describe the topology of the organization. FS-DFM, when applied to the generation space, describes the topology of the transformation. At the intersection of the two, Transition Computing is born, an architecture where data, model, and knowledge coexist as states of the same discrete field. Thus, by revisiting the concepts presented in 2017 and placing them in dialogue with the most recent formulations of 2025, we reaffirm that the true computational revolution lies not only in the speed of calculation, but in understanding the leap. The crossing, once seen as inevitable, becomes optional; the path, once traveled, is now understood. And it is at this moment, where the determinism of SRMs meets the flow of FS-DFM, that the new paradigm of computational intelligence is born: an intelligence that doesn't walk, but transits. Reference [1] França, C, R 2017 Heru Search Method—Unique in the World that Uses Unprecedented Mathematical Formulas and Replaces the Binary Tree Breaking Various Paradigms Like 0(log n), American Jornal of Computational Mathematics - DOI 10.4236/ajcm.2017.71003 [2] França, C. (2025). Beyond Binary Trees: A Paradigm Shift in Data Search Efficiency via SRMs-Based Structures. Zenodo. https://doi.org/10.5281/zenodo.15602787 [3] França, C. R. (2025). Método de criptografia Heru Technologies: único do mundo que utiliza fórmulas matemáticas inéditas autorais e que criptografa com perturbações binárias. Zenodo. https://doi.org/10.5281/zenodo.15653585