Zhuldyz Talasbayeva

Prof. Zhuldyz Talasbayeva

orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
5 November 2026 in Information
Verified
Journal Article
Open access
PDF file
YTYerkezhan Tleukulova
Yerkezhan Tleukulova
Madina MansurovaMadina Mansurova
Madina Mansurova
ENElmira Shodmonovna Nazirova
Elmira Shodmonovna Nazirova
SKSabira Kulchimbayeva
Sabira Kulchimbayeva
ZTZhuldyz Talasbayeva
Zhuldyz Talasbayeva
Talshyn SarsembayevaTalshyn Sarsembayeva
Talshyn Sarsembayeva
Perinatal depression affects a large share of women during pregnancy and after childbirth, yet access to perinatal mental health care in Kazakhstan is limited. This paper presents MamaCare, a multi-agent platform based on open-weight Qwen3.5 large language models (LLMs) that provides conversational support, administers standardized screening instruments, and triages red-flag situations in Kazakh, Russian, and English. Two physician experts rated 286 single- and multi-turn sessions against a rubric adapted from HealthBench and ThReadMed-QA; mean scores for the accuracy, safety, and communication criteria ranged from 4.45 to 4.93 out of 5, and the lowest-scoring criterion was turn-level degradation (3.71). The platform was then compared with five prompt-engineered configurations of its own base model on 120 stratified cases, and every automated judgment was repeated with three judges from three model families: Qwen3.8-27B, Gemma-3-27B, and a Kazakh-adapted Llama-3.1-8B. In red-flag triage, MamaCare had the lowest false-alarm rate of all six configurations under all three judges and against physician urgency labels for the 58 cases on which any judge raised an alarm, and a prompt-only safety configuration escalated routine questions 3.6 to 4.2 times more often; agreement between judges was substantial for urgency labelling (Cohen’s κ = 0.67 and 0.79) and for escalation detection (κ = 0.78 and 0.47). Pairwise preference did not survive the same test (κ = 0.30, with 15.3% to 45.7% of verdicts reversing when the answer order was swapped) and is therefore reported as judge-dependent evidence rather than as a result. Retention of facts from earlier turns was the platform’s weakest property (71.8%); adding a structured patient-state handoff between agents raised it to 97.4% under both judges that measure it reliably, without increasing false alarms and with shorter answers. On a new holdout set of 15 parallel multi-turn dialogues in the three languages, written after the platform was fixed, physician-rated fact retention over the 10 completed dialogues was 0.85 (95% CI 0.74–0.91), with no difference between languages; the main multi-turn errors were information carried over from the user profile and replies truncated by server time-outs.

23 December 2019 in Journal of Mathematics, Mechanics and Computer Science
Verified
Journal Article
Open access
PDF file
ZNZ. Nurpeyis
Z. Nurpeyis
ZTZhuldyz Talasbayeva
Zhuldyz Talasbayeva
AMA. D. Mazhitova
A. D. Mazhitova
Предлагаемая работа посвящена выделению и изучению многомерных сетей, конструктив-но связанных распределением. В первоначальном подходе к выделению сетей существенноиспользуется вектор средней кривизны распределения. Поэтому такое выделение осуществи-мо лишь в метрических пространствах (в работе этот вопрос исследуется в евклидовом n-пространстве). В статье исследуются условия существования канонических распределенийплоскостей, принадлежащих касательной плоскости поверхности евклидова пространства.В данной статье введено понятие параллельного переноса площадки вдоль интегральныхкривых однораспределения. Доказано утверждение о том, что векторные поля коллинеарнытогда и только тогда, когда геометрический объект тензор кривизны является нулевым.Выведены дифференциальные уравнения однораспределения и найдено необходимое и до-статочное условие, для того чтобы геодезическая линия однораспределения была плоской, атакже найдены условия, при которых интегральные кривые однораспределения являются ли-ниями кривизны относительно однораспределения. Доказано утверждение о том, что линиябудет геодезической, тогда и только тогда, когда ее главные нормали совпадают с нормалямиповерхности, на который эта линия расположена. Получены дифференциальные уравнениягедезической плоской линии.

17 October 2026 in Robotics
Verified
Journal Article
Open access
PDF file
Nurgali KadyrbekNurgali Kadyrbek
Nurgali Kadyrbek
Madina MansurovaMadina Mansurova
Madina Mansurova
ZTZhuldyz Talasbayeva
Zhuldyz Talasbayeva
ZZZhadyra Zhumasheva
Zhadyra Zhumasheva
Task failure alone cannot distinguish limited robot execution from incorrect target selection. We evaluate released SmolVLA and π0 policies in VLABench using family-matched native controls, same-state trusted and null instructions, online symmetric stopping and a separate retention check. The main experiment contains 1860 attempts under three inference seeds; a command-duration follow-up adds 420. Native completion rises from 0/210 and 1/210 at 6 simulated seconds to 73/210 and 101/210 at 20 s, respectively, correcting a short evaluation cutoff. Reviewer-requested secondary analysis of 40 benchmark-derived paired scenes gives a symmetric request-swap contrast of 10.00 percentage points for π0 (scene-bootstrap 95% CI [3.33, 17.08]) and zero for SmolVLA ([−4.17, 4.58]), yet requested choice under switched instructions is only 20/47 (42.6%) and 9/22 (40.9%) among goal-reaching attempts, respectively. Across all 60 nominal fixtures, the π0 advantage over blank input decomposes into 11.39 points of coverage and 3.33 points of net-choice contribution ([1.11, 5.69]). Its shuffled comparison warrants caution: the original adjusted p=0.0481 is borderline; 10/40 benchmark donors name a candidate, and all 20 constructed donors equal request A. Holding each command for two environment steps rather than one reduces native completion to 39/210 and 55/210 at the same 20 s budget while also changing command count and replanning time. The results demonstrate configuration-sensitive execution and a bounded instruction effect, not reliable target switching, architecture superiority or sim-to-real transfer.

Verified
Journal Article
Open access
PDF file
Nurgali KadyrbekNurgali Kadyrbek
Nurgali Kadyrbek
Madina MansurovaMadina Mansurova
Madina Mansurova
ZTZhuldyz Talasbayeva
Zhuldyz Talasbayeva
ZZZhadyra Zhumasheva
Zhadyra Zhumasheva
We learn compatible destination and owner views in bounded, transcript-free records. Structural source/type/version addressing selects the first relation; authoritative, requested-attribute filtering reduces ordinary candidates to a mean of 3.28. Learned matching selects the second record and retrieves its value. Three seeds of a 24.35-million-parameter multitask model achieve 1.0000 one-hop, identity, and two-hop accuracy on shared sealed confirmation rows. Relation and value erasures collapse accuracy; counterfactual replacements redirect answers. A separately preregistered extension reserves disjoint compositional names across training, development, and sealed confirmation while matching tokenization. After 512 unrelated writes, challenge identity and two-hop accuracy are 0.9990–1.0000 with 16 or 32 candidates; ordinary held-out-name two-hop accuracy is 0.9328–0.9567. Ordinary-only training also solves the original naming distribution, so explicit challenge examples are not necessary there. Cutting direct writer credit reduces challenge identity. Complete replay succeeds, unlike last-32-event replay. Earlier protected bindings fail two-hop retrieval; generic recurrence fails in its tested multitask regime but a semantic-only GRU acquires shorter-delay memory. A separate zero-shot name/tokenization shift remains unsuccessful. These results establish causally functional two-access memory and controlled compositional-name transfer, not unrestricted language understanding or general reasoning. Several other families remain near chance.

24 September 2026 in Journal of Mathematics, Mechanics and Computer Science
Verified
Journal Article
Open access
PDF file
ZTZhuldyz Talasbayeva
Zhuldyz Talasbayeva
Algorithmic learning of algebraic structures connects computability theory with artificial intelligence. Classical inductive inference mainly deals with formal languages and recursive functions. In contrast, analyzing vector spaces allows us to explore the theoretical limits of automated scientific discovery. This study investigates the algorithmic learnability of subspace families in computable vector spaces over a computable field. We focus on the fundamental family of all finite-dimensional subspaces and examine whether learnability is an isomorphism invariant. By combining constructive model theory with Gold’s paradigm of identification in the limit, our results identify the exact recursion-theoretic barriers to learning algebraic invariants. We use the classical model of explanatory learning in the limit from positive data (Ex-learning), where Turing machines act as learners. By applying canonical index numberings of finite sets and effective quotient space constructions, we establish two fundamental theorems. Our first result shows that if a computable vector space has a decidable linear dependence relation, the family of all its finite dimensional subspaces is Ex-learnable.. Second, we construct a computable vector space where the relation of linear dependence is undecidable. By using an effective priority-style diagonalization argument, we prove that the family of finite-dimensional subspaces in this space fails to be Ex learnable. These results prove that algorithmic learnability is not an isomorphism invariant of an abstract algebraic structure. Instead, it is strictly governed by the algorithmic complexity (Turing degree) of the linear dependence predicate. This finding provides a theoretical foundation for automated theorem proving and neuro-symbolic AI systems.