arXiv subjects

Viktoria Rojkova

Publications and source records attributed to Viktoria Rojkova.

9 recordsLinked to original sources

arXiv2026

Context by Distinct Information: An Auditable Dirichlet-Process Working Memory for Long, Redundant Context Streams

Context engineering decides what information a model carries forward, and current designs meter it in tokens: compressing the past into a bounded recurrent state, keeping a key-value entry for every token, or imposing a fixed budget through a window or eviction rule. All three make the token the unit of memory even when the stream is redundant and the task depends on the distinct information it carries. Building on a companion mechanism paper that opens a cache slot only when an incoming key is novel, so memory scales with the number of distinct items rather than tokens, we develop that allocate-on-novelty cache as a working-memory component and organize context by how a task depends on the past: recall-carried information belongs in a content-addressed novelty cache, summary-carried information in a recurrent state, and locality-carried information in a recency window. The claim is empirical and bounded. On a matched character-level control, novelty-gated attention reaches full-attention performance while attending to about half the tokens, and coupling the cache with a state-space summary matches full-attention coupling at that reduced cost; the advantage grows as context lengthens, while a sliding window is preferable on short, locality-dominated spans. On next-code prediction over synthetic Medicare claims the coupled component leads full attention and every fixed-budget eviction policy at a thousand-event horizon, whereas cost forecasting over the same stream is summary-carried and the cache is neutral. The retained memory is an inspectable table of templates, codes, drugs, or places rather than an opaque state. The experiments are small-scale and use only public data; they establish the primitive that context can scale with distinct information rather than tokens, in a working memory that is content-addressable and auditable.

cs.LG↗

arXiv2026

Knowledge as Orbit: Finite Collections as Phases of an Exactly Periodic Latent Generator

Finite knowledge is usually stored extensionally, one code or vector per item. We ask whether a finite collection can instead be stored intensionally, as the decoded orbit of one compact law that returns exactly to its start. For X objects, we encode item i as the i-th phase of a fixed rotation in a learned latent space and decode all phases with a shared network; the latent advances through a bank of rotations at integer harmonics of the cycle, a real discrete Fourier operator, so that R^X equals the identity and exact closure is guaranteed rather than learned. Images are a controlled carrier; looping video is the case where the phase order is the content's own temporal structure. Holding the decoder fixed and varying only the operator, a general learned operator diverges, a norm-preserving but non-periodic one degrades around the loop, and the exactly periodic operator is flat; on real images the gap widens. Capacity is then the decoder's budget: dense decoders carry a structural overhead per crisp image that no size reconciles with compression, while a small convolutional decoder on objects that share a manifold reaches crisp and compressed. A codebook control shows the generative law is free in reconstruction terms while multiplying the latent store many-fold. On seven benchmark clips, against a matched frame-index baseline, the cycle reaches equal or better fidelity at equal parameters while wrapping at machine precision, where the baseline leaves a visible seam; pinning the baseline's frequencies to loop harmonics closes its seam too, confirming that exact periodicity is the operative constraint. Finite cyclic knowledge can be stored as dynamics rather than independent instances, with exact recurrence supplied by algebra and content by a shared decoder.

cs.LG↗

arXiv2026

Remembering Distinct Items, Not Tokens: A Learnable Dirichlet-Process Cache Between State-Space Models and Attention

Fixed-state sequence models compress an unbounded past into a bounded state, which caps their associative recall at roughly the state dimension; attention escapes the cap by keeping a key-value entry for every token, at quadratic compute and a cache that grows with the sequence. We study the middle ground: a sparse cache that allocates a slot only when an input is novel, so its size tracks the number of distinct items rather than the number of tokens. The allocation rule is the DP-means clustering rule, the small-variance limit of a Dirichlet-process mixture, used not as latent-variable inference but as the key-value memory operator for a deep recurrent backbone. We develop it in two forms, a static cache with a fixed concentration and a surprise-adaptive variant whose concentration follows the recent novelty rate. On a controlled associative-recall benchmark with redundancy we show that the cache matches full-attention recall while storing only the distinct items, that it dominates a fixed-budget eviction cache on the recall-versus-size frontier, and that on a state-space backbone it answers both a recall query and a long-range aggregate at the lowest memory of any model tested. The allocation is learnable end to end: a two-parameter novelty-threshold gate trained on the task loss alone recovers the rule exactly, whereas an over-parameterized gate fails, so the operative ingredient is the inductive bias rather than capacity. The evidence is a family of controlled mechanism studies at modest scale, with the distinct-items property confirmed on four real streams (recommendation, systems logs, clinical events, and insurance claims); a real-backbone, real-corpus language validation is pursued in a companion study.

cs.LG↗

arXiv2026

A Stationarity-and-Coupling Criterion for Training-Free Time-Lagged Spectral Embeddings of Multivariate Time Series

We study training-free fixed-length descriptors for multivariate time series and ask not merely whether such a descriptor performs well, but when it can be expected to work at all. Our object of study is $D(τ)$, built from a time-lagged correlation matrix truncated at the Marchenko-Pastur edge so that only signal-bearing eigenvalues survive and classified by cosine similarity to class centroids with zero learned parameters. The central contribution is not the descriptor but a falsifiable applicability criterion for it. Working from a stationary Gaussian VAR(1) model, we argue that $D(τ)$ separates two classes when the signals are approximately stationary and the class information lives in their cross-channel temporal coupling rather than in marginal per-channel power. We derive, semi-formally, three consequences: a distinguishability condition, why the static ($τ=0$) covariance collapses to chance, and why a stationary but power-discriminated paradigm defeats the descriptor. The criterion is operational: a two-part pre-flight test -- an augmented Dickey-Fuller stationarity check and a power-baseline saturation check -- predicts applicability before any training. We validate both halves on a mixed assortment. On four paradigms that satisfy the criterion (Sleep-EDF, BCI-IV-2a, MIT-BIH, ESC-50) the descriptor is competitive with strong baselines at a fraction of their cost, reaching $88.5\pm4.5\%$ under 20-subject leave-one-subject-out on Sleep-EDF on a single CPU thread. On three that violate it -- non-stationary ERPs, and financial-volatility and wearable-stress regimes that are power-discriminated -- it fails exactly as the pre-flight predicts, and these negatives are the more informative half. We are explicit that $D(τ)$ is not the most accurate representation; its value is a compact, training-free embedding whose domain of validity is known in advance.

cs.LG↗

arXiv2025

From Paper to Structured JSON: An Agentic Workflow for Compliant BMR Digital Transformation

Pharmaceutical manufacturers generate thousands of batch manufacturing records (BMRs) each year under FDA 21 CFR Part 211 and EU GMP rules. These long documents combine tables, calculations, images, and handwritten notes, and are usually digitized by hand with hours of expert review per record. We present an AI workflow that converts unstructured BMRs into structured JSON using token based chunking, parallel large language model extraction, and a fixed schema that covers 11 content types while preserving the original Group-Phase-Step hierarchy. The system applies three layers of validation (JSON syntax, structural integrity of classes and references, and pharmaceutical compliance checks aligned with GMP) and reports coverage metrics for text, tables, images, and calculations. On three real BMRs between 15 and 66 pages, it achieves composite confidence scores in the low to high eighties while reducing processing time from hours to minutes on a single GPU. This enables practical, human in the loop BMR digitization at scale and unlocks historical manufacturing data for downstream analysis.

cs.DL↗

arXiv2025

RAGulating Compliance: A Multi-Agent Knowledge Graph for Regulatory QA

Regulatory compliance question answering (QA) requires precise, verifiable information, and domain-specific expertise, posing challenges for Large Language Models (LLMs). In this work, we present a novel multi-agent framework that integrates a Knowledge Graph (KG) of Regulatory triplets with Retrieval-Augmented Generation (RAG) to address these demands. First, agents build and maintain an ontology-free KG by extracting subject--predicate--object (SPO) triplets from regulatory documents and systematically cleaning, normalizing, deduplicating, and updating them. Second, these triplets are embedded and stored along with their corresponding textual sections and metadata in a single enriched vector database, allowing for both graph-based reasoning and efficient information retrieval. Third, an orchestrated agent pipeline leverages triplet-level retrieval for question answering, ensuring high semantic alignment between user queries and the factual "who-did-what-to-whom" core captured by the graph. Our hybrid system outperforms conventional methods in complex regulatory queries, ensuring factual correctness with embedded triplets, enabling traceability through a unified vector database, and enhancing understanding through subgraph visualization, providing a robust foundation for compliance-driven and broader audit-focused applications.

cs.AI↗

arXiv2025

Think Inside the JSON: Reinforcement Strategy for Strict LLM Schema Adherence

In this paper, we address the challenge of enforcing strict schema adherence in large language model (LLM) generation by leveraging LLM reasoning capabilities. Building on the DeepSeek R1 reinforcement learning framework, our approach trains structured reasoning skills of a 1.5B parameter model through a novel pipeline that combines synthetic reasoning dataset construction with custom reward functions under Group Relative Policy Optimization (GRPO). Specifically, we first perform R1 reinforcement learning on a 20K sample unstructured-to-structured dataset, mirroring the original DeepSeek R1 methods, to establish core reasoning abilities. Subsequently, we performed supervised fine-tuning on a separate 10K reasoning sample dataset, focusing on refining schema adherence for downstream tasks. Despite the relatively modest training scope, requiring approximately 20 hours on an 8xH100 GPU cluster for GRPO training and 3 hours on 1xA100 for SFT, our model demonstrates robust performance in enforcing schema consistency. We compare our ThinkJSON approach against the original DeepSeek R1 (671B), distilled versions of DeepSeek R1 (Qwen-1.5B and Qwen-7B), and Gemini 2.0 Flash (70B), showcasing its effectiveness in real-world applications. Our results underscore the practical utility of a resource-efficient framework for schema-constrained text generation.

cs.CL↗

arXiv2007

Delayed Correlations in Inter-Domain Network Traffic

To observe the evolution of network traffic correlations we analyze the eigenvalue spectra and eigenvectors statistics of delayed correlation matrices of network traffic counts time series. Delayed correlation matrix D is composed of the correlations between one variable in the multivariable time series and another at a time delay τ. Inverse participation ratio (IPR) of eigenvectors of D deviates substantially from the IPR of eigenvectors of the equal time correlation matrix C. We relate this finding to the localization and discuss its importance for network congestion control. The time-lagged correlation pattern between network time series is preserved over a long time, up to 100τ, where τ=300 sec. The largest eigenvalue λ_{max} of D and the corresponding IPR oscillate with two characteristic periods of 3τand 6τ. The existence of delayed correlations between network time series fits well into the long range dependence (LRD) property of the network traffic. The ability to monitor and control the long memory processes is crucial since they impact the network performance. Injecting the random traffic counts between non-randomly correlated time series, we were able to break the picture of periodicity of λ_{max}. In addition, we investigated influence of the periodic injections on both largest eigenvalue and the IPR, and addressed relevance of these indicators for the LRD and self-similarity of the network traffic.

cs.NI↗

arXiv2007

Analysis of Inter-Domain Traffic Correlations: Random Matrix Theory Approach

The traffic behavior of University of Louisville network with the interconnected backbone routers and the number of Virtual Local Area Network (VLAN) subnets is investigated using the Random Matrix Theory (RMT) approach. We employ the system of equal interval time series of traffic counts at all router to router and router to subnet connections as a representation of the inter-VLAN traffic. The cross-correlation matrix C of the traffic rate changes between different traffic time series is calculated and tested against null-hypothesis of random interactions. The majority of the eigenvalues λ_{i} of matrix C fall within the bounds predicted by the RMT for the eigenvalues of random correlation matrices. The distribution of eigenvalues and eigenvectors outside of the RMT bounds displays prominent and systematic deviations from the RMT predictions. Moreover, these deviations are stable in time. The method we use provides a unique possibility to accomplish three concurrent tasks of traffic analysis. The method verifies the uncongested state of the network, by establishing the profile of random interactions. It recognizes the system-specific large-scale interactions, by establishing the profile of stable in time non-random interactions. Finally, by looking into the eigenstatistics we are able to detect and allocate anomalies of network traffic interactions.

cs.NI↗