https://emanuelshirbint.github.io/research/ Independent scholarly research archive and author publication platform. Verified research works with publication and peer-review status. en 2026-08-18 https://emanuelshirbint.github.io/research/works/bounded-recursive-self-improvement/ https://emanuelshirbint.github.io/research/works/bounded-recursive-self-improvement/ Emanuel Shirbint Version 2 preprint · Not peer-reviewed · Governed local experiment 2026-08-17 A bounded seven-cycle scaffolding-level experiment with separate human promotion authority. The public record reports four promoted states, one retained failed transition, and two NO_CHANGE outcomes; it does not establish AGI, unbounded improvement, autonomous promotion, permanent saturation, or external production validation. https://emanuelshirbint.github.io/research/works/reinforcement-learning-contour/ https://emanuelshirbint.github.io/research/works/reinforcement-learning-contour/ Emanuel Shirbint Not peer-reviewed 2026-08-10 Crossref and Europe PMC verify a Version 2 preprint record posted on 10 August 2026 with the same title and sole author as Version 1. The record remains non-peer-reviewed; the archived local PDF is Version 1. https://emanuelshirbint.github.io/research/works/institutional-volition-systemic-error/ https://emanuelshirbint.github.io/research/works/institutional-volition-systemic-error/ Emanuel Shirbint Not peer-reviewed 2026-07-15 Evaluation of artificial intelligence concentrates on local performance — accuracy, calibration, robustness — while the upstream normative and volitional architecture that fixes what a system is for remains largely unexamined by those metrics. Existing scholarship has established that bias, proxy variables, problem formulation, measurement, alignment, and governance are value-laden; what remains i https://emanuelshirbint.github.io/research/works/measuring-what-survives/ https://emanuelshirbint.github.io/research/works/measuring-what-survives/ Emanuel Shirbint, Alexander Rybalov Not peer-reviewed 2026-06-16 When a safety intervention lowers an agent’s attack-success rate, two different things may have happened: the intervention removed a load-bearing failure route, or the agent learned the measured attacks while the route survived. Frequency metrics are necessary, but they do not by themselves identify which behavioural structures persist as adversarial pressure rises. We propose a measurement framew https://emanuelshirbint.github.io/research/works/topological-robustness-llm-agents/ https://emanuelshirbint.github.io/research/works/topological-robustness-llm-agents/ Emanuel Shirbint, Alexander Rybalov Not peer-reviewed 2026-06-09 Adversarial evaluation of large-language-model agents is usually expressed as event counts: attack-success rate, policy-violation rate, refusal-failure rate, pass percentage, or robustness score. These measures are necessary, but they do not show whether a safety intervention removed a load-bearing failure mode or merely suppressed the measured cases. This paper presents a model-agnostic framework https://emanuelshirbint.github.io/research/works/human-digital-twins-self-knowledge/ https://emanuelshirbint.github.io/research/works/human-digital-twins-self-knowledge/ Emanuel Shirbint, Alexander Rybalov Not peer-reviewed 2026-06-02 Large language models embedded in person-modelling systems, and human beings reflecting on themselves, share a structural vulnerability: both can generate fluent, narratively persuasive, yet abductively unsound accounts of a person. Building on a governed abductive architecture developed for medical patient digital twins, this paper argues that personality should be modelled neither as a stable se https://emanuelshirbint.github.io/research/works/medical-digital-twins-sound-abduction/ https://emanuelshirbint.github.io/research/works/medical-digital-twins-sound-abduction/ Emanuel Shirbint, Alexander Rybalov Not peer-reviewed 2026-05-12 Large language models (LLMs) embedded in medical Patient Digital Twins (PDTs) exhibit a systemic vulnerability: they can generate fluent, narratively persuasive, yet abductively unsound clinical explanations. In this context, abductive soundness means that an explanation preserves mechanistic plausibility, temporal coherence, explicit handling of missing premises, and sensitivity to counter-eviden https://emanuelshirbint.github.io/research/works/totalspineseg/ https://emanuelshirbint.github.io/research/works/totalspineseg/ Yehuda Warszawer, Nathan Molinier, Jan Valošek, Pierre-Louis Benveniste, Sandrine Bédard, Emanuel Shirbint, Feroze B. Mohamed, Charidimos Tsagkas, Shannon Kolind, Larry Lynd, Jiwon Oh, Alexandre Prat, Roger Tam, Anthony Traboulsee, Scott Patten, Lisa Eunyoung Lee, Anat Achiron, Julien Cohen-Adad Preprint; associated conference work is not a journal peer-review record 2025-03-16 Accurate segmentation and labeling of spinal structures are essential for diagnosing and managing spinal pathologies. TotalSpineSeg uses cascaded nnU-Net models and landmark-guided iterative labeling to segment vertebrae, intervertebral discs, spinal cord, and spinal canal across MRI contrasts. https://emanuelshirbint.github.io/research/works/finger-dexterity-biosensing/ https://emanuelshirbint.github.io/research/works/finger-dexterity-biosensing/ Michal Greenberg Abrahami, Yehuda Warszawer, Alon Kalron, Emanuel Shirbint, Maria Didikin, Anat Achiron Peer-reviewed journal article 2025-01-01 Objective Accurate finger function assessment is crucial for monitoring the performance of daily hand activities. However, specialized digital applications are lacking for evaluating various finger tasks. This study aims to develop a custom digital biosensing application to assess finger dexterity. Methods We developed a digital biosensing application compatible with smartphones and tablets that e https://emanuelshirbint.github.io/research/works/ppms-brain-volume/ https://emanuelshirbint.github.io/research/works/ppms-brain-volume/ Yehuda Warszawer, Michael Gurevich, Ariel Kerpel, Sapir Dreyer Alster, Yael Nissan, Emanuel Shirbint, Chen Hoffmann, Anat Achiron Peer-reviewed journal article 2024-04-13 Purpose Detection and prediction of the rate of brain volume loss with age is a significant unmet need in patients with primary progressive multiple sclerosis (PPMS). In this study we construct detailed brain volume maps for PPMS patients. These maps compare age-related changes in both cortical and sub-cortical regions with those in healthy individuals. Methods We conducted retrospective analyses https://emanuelshirbint.github.io/research/works/ms-third-booster/ https://emanuelshirbint.github.io/research/works/ms-third-booster/ Sapir Dreyer-Alster, Shay Menascu, Mathilda Mandel, Emanuel Shirbint, David Magalashvili, Mark Dolev, Shlomo Flechter, Uri Givon, Diana Guber, Yael Stern, Shmuel Miron, Michael Polliack, Rina Falb, Polina Sonis, Michael Gurevich, Anat Achiron Peer-reviewed journal article 2022-03-01 Background As immunity against SARS-COV-2 wanes following first and second doses of vaccination, a third dose is administered in several countries around the world. Similarly to the first doses, risks related to vaccination and humoral immune response in patients with multiple sclerosis (MS) need to be assessed. Objective Characterize safety and humoral immune response following the third dose of https://emanuelshirbint.github.io/research/works/ms-vaccination-february-2021/ https://emanuelshirbint.github.io/research/works/ms-vaccination-february-2021/ Anat Achiron, Mark Dolev, Shay Menascu, Daniela-Noa Zohar, Sapir Dreyer-Alster, Shmuel Miron, Emanuel Shirbint, David Magalashvili, Shlomo Flechter, Uri Givon, Diana Guber, Yael Stern, Michael Polliack, Rina Falb, Michael Gurevich Peer-reviewed journal article 2021-04-15 Background Since vaccination against coronavirus disease 2019 (COVID-19) became available, risks related to vaccinating patients with multiple sclerosis (MS) need to be carefully assessed. Objective Characterize safety and occurrence of immediate relapses following COVID-19 vaccination in a large cohort of MS patients. Methods We assessed the safety of BNT162b2 COVID-19 vaccination in adult MS pat