"úÁÐÉÓËÉ ÎÁÕÞÎÙÈ ÓÅÍÉÎÁÒÏ× ðïíé"

• ðÒÅÄÉÓÌÏ×ÉÅ ÏÔ ÒÅÄÁËÔÏÒÁ ....... 5 ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• áÌÅËÓÅÅ×ÓËÁÑ é., áÒÈÉÐÅÎËÏ ë. äÖÅÊÌÂÒÅÊË ÂÏÌØÛÉÈ ÑÚÙËÏ×ÙÈ ÍÏÄÅÌÅÊ Ó ÐÏÍÏÝØÀ ÍÑÇËÏÇÏ ÐÅÒÅÆÒÁÚÉÒÏ×ÁÎÉÑ ×ÒÅÄÏÎÏÓÎÙÈ ÚÁÐÒÏÓÏ× ....... 7 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• áÎÇÅÎÉ ç. ü., åÒÛÏ× ÷. á. ëÌÉÅÎÔÏÏÂÕÓÌÏ×ÌÅÎÎÏÅ ÓÐÅËÕÌÑÔÉ×ÎÏÅ ÄÅËÏÄÉÒÏ×ÁÎÉÅ ÄÌÑ ÂÏÌØÛÉÈ ÑÚÙËÏ×ÙÈ ÍÏÄÅÌÅÊ × ÍÕÌØÔÉÁÒÅÎÄÎÙÈ ÓÅÒ×ÉÓÁÈ ....... 25 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• âÁÒÙÛÎÉËÏ× ï., áÌÅËÓÅÅ× á. í., îÉËÏÌÅÎËÏ ó. é. Query2Diagram: ÏÔ×ÅÔÙ ÎÁ ÚÁÐÒÏÓÙ ÐÏ ËÏÄÕ Ó ÐÏÍÏÝØÀ UML-ÄÉÁÇÒÁÍÍ ....... 63 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• âÕÔÁËÏ× î., ÷ÁÈÒÕÛÅ× á., èÏÄÏÒÞÅÎËÏ í., úÁÈÁÒÏ×Á á., òÙÖËÏ× á., ðÌÏÓÓËÁÑ ï., çÁÊÎÅÔÄÉÎÏ× á., äÁÍÄÉÎÏ× ò., áÌÅËÓÁÎÄÒÏ× ä., îÁÓÏÎÏ× ä., óÁ×ÞÅÎËÏ í. üÆÆÅËÔÉ×ÎÏÅ ÐÌÁÎÉÒÏ×ÁÎÉÅ ÒÁÓÐÒÅÄÅÌ£ÎÎÙÈ AutoML-ËÏÎ×ÅÊÅÒÏ× Ó ÕÞ£ÔÏÍ ÄÅÄÌÁÊÎÏ× ....... 87 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• äÅÍÅÎÔØÅ× à. é., éÇÎÁÔØÅ× á. á., óÉÄÅÌØÎÉË ÷. á., óÍÁÌØ á. ÷., õÛÁËÏ× í. ó. îÏ×ÙÅ ÒÅÚÕÌØÔÁÔÙ Ï ÐÏÌÕÄÕÐÌÅËÓÎÏÊ ËÏÍÍÕÎÉËÁÃÉÏÎÎÏÊ ÓÌÏÖÎÏÓÔÉ ....... 100 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• åÌÉÓÔÒÁÔÏ× ó., ðÏÄÉ×ÉÌÏ× á., àÖÁËÏ× ô., ÷ÅÔÒÏ× ä. îÁÐÒÁ×ÌÅÎÎÙÊ ÛÕÍ É ÎÅÑ×ÎÏÅ ÓÍÅÝÅÎÉÅ × ÍÅÔÏÄÅ Lion ....... 134 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• çÁÌÉÍÚÑÎÏ×Á ä., çÏÒÏ×ÁÑ ó., ÷ÉÈÏÒÅ× ä., ñÍÝÉËÏ× é. ð., öÅÍÞÕÖÉÎÁ å. CleanComedy: ÓÏÚÄÁÎÉÅ ÄÒÕÖÅÌÀÂÎÏÇÏ ÀÍÏÒÁ Ó ÐÏÍÏÝØÀ ÇÅÎÅÒÁÔÉ×ÎÙÈ ÍÅÔÏÄÏ× ....... 153 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• çÁÒÉÐÏ× ò., úÁÂÁÛÔÁ á., æÉÌØÞÅÎËÏ× á. áÒÈÉÔÅËÔÕÒÁ ÇÌÕÂÏËÏÇÏ ÏÂÕÞÅÎÉÑ ÄÌÑ ÏÂÒÁÂÏÔËÉ ÔÁÂÌÉÞÎÙÈ ÄÁÎÎÙÈ Ó ÃÅÌØÀ ÉÚ×ÌÅÞÅÎÉÑ ÍÅÔÁ-ÐÒÉÚÎÁËÏ× ....... 178 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• éÏ× é. ì., îÉËÉÔÉÎ î. ï. ïÐÔÉÍÉÚÁÃÉÑ ËÏÎ×ÅÊÅÒÁ ËÏÎÓÔÒÕÉÒÏ×ÁÎÉÑ ÐÒÉÚÎÁËÏ× × ÐÒÏÃÅÓÓÅ AutoML Ó ÉÓÐÏÌØÚÏ×ÁÎÉÅÍ ÂÏÌØÛÉÈ ÑÚÙËÏ×ÙÈ ÍÏÄÅÌÅÊ ....... 191 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• ìÁÜÔÉÎ á. á., ðÕÌÆÅÒ â., óÍÉÒÎÏ× ó. ë. éÅÒÁÒÈÉÞÅÓËÁÑ ÉÎÉÃÉÁÌÉÚÁÃÉÑ ×ÒÅÍÅÎÎÙÈ ÍÁÓÛÔÁÂÏ× × Mamba-2 ÄÌÑ ÍÎÏÇÏÍÁÓÛÔÁÂÎÏÇÏ ÍÏÄÅÌÉÒÏ×ÁÎÉÑ ÐÏÓÌÅÄÏ×ÁÔÅÌØÎÏÓÔÅÊ ....... 223 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• íÏÌÏÄÅÃËÉÈ é. á., íÁÌÙÛÅ× ë. ÷., íÉÒÇÁÌÅÅ× í. ò., âÏÇÁÔÙÒ£× å. î., ÷ÁÔÏÌÉÎ ä. ó. íÅÔÏÄÏÌÏÇÉÑ ÓÂÏÒÁ É ÁÎÁÌÉÚÁ ÓÕÂßÅËÔÉ×ÎÏÊ ÚÁÍÅÔÎÏÓÔÉ ÁÒÔÅÆÁËÔÏ× ÎÅÊÒÏÓÅÔÅ×ÙÈ ÍÏÄÅÌÅÊ ÓÕÐÅÒÒÁÚÒÅÛÅÎÉÑ ....... 235 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• îÉËÏÌÅÎËÏ ó. é. íÅÔÏÄÙ ÆÉÌØÔÒÁÃÉÉ É ÏÃÅÎËÉ ÄÁÎÎÙÈ ÄÌÑ ÂÏÌØÛÉÈ ÑÚÙËÏ×ÙÈ ÍÏÄÅÌÅÊ ....... 254 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• îÉËÏÌÅÎËÏ ó. é. æÕÎËÃÉÉ ×ÌÉÑÎÉÑ ÄÁÎÎÙÈ ÄÌÑ ÂÏÌØÛÉÈ ÑÚÙËÏ×ÙÈ ÍÏÄÅÌÅÊ ....... 280 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• ïÂÌÏ×ÁÔÎÙÊ ò., ëÕÌÅÛÏ×Á á., ðÏÌÅ× ë., úÁÊÃÅ× á. òÁÓÐÏÚÎÁ×ÁÎÉÅ ÇÁÌÌÀÃÉÎÁÃÉÊ × ÂÏÌØÛÉÈ ÑÚÙËÏ×ÙÈ ÍÏÄÅÌÑÈ Ó RAG ÎÁ ÏÓÎÏ×Å ×ÅÒÏÑÔÎÏÓÔÎÙÈ ÒÁÓÓÔÏÑÎÉÊ ....... 327 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• ïÒÌÏ× å., íÁÌÙÈ ÷. ðÏÖÁÌÕÊÓÔÁ, ÐÒÏÇÏÌÏÓÕÊÔÅ: ÒÁÎÖÉÒÏ×ÁÎÉÅ ÏÔ×ÅÔÏ× ÎÁ Stack Overflow × ÔÅËÓÔÏ×ÏÊ ÐÏÓÔÁÎÏ×ËÅ ....... 350 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• ûÕÍÉÃËÁÑ å. á., ÷ÁÔÏÌÉÎ ä. ó., áÎÃÉÆÅÒÏ×Á á. ÷. éÓÐÏÌØÚÏ×ÁÎÉÅ ÓÔÏÈÁÓÔÉÞÅÓËÉÈ ÕÑÚ×ÉÍÏÓÔÅÊ ÒÁÎÄÏÍÉÚÉÒÏ×ÁÎÎÏÇÏ ÓÇÌÁÖÉ×ÁÎÉÑ × ÒÅÇÒÅÓÓÉÏÎÎÙÈ ÍÏÄÅÌÑÈ: ÁÎÁÌÉÚ ÁÄÁÐÔÉ×ÎÙÈ ÁÔÁË É ÜÍÐÉÒÉÞÅÓËÉÅ ÏÇÒÁÎÉÞÅÎÉÑ ....... 381 áÎÎÏÔÁÃÉÑ ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)
• òÅÆÅÒÁÔÙ ....... 398 ðÏÌÎÙÊ ÔÅËÓÔ(.pdf)

---


Contents 
• Preface ....... 5
• Alekseevskaia I., Arkhipenko K. Jailbreaking large language models by soft-paraphrasing the malicious prompts ....... 7
• Angeni G. E., Ershov V. A. Client-conditioned speculative decoding for large Language models in multi-tenant services ....... 25
• Baryshnikov O., Alekseev A. M., Nikolenko S. I. Query2Diagram: answering developer queries with UML diagrams ....... 63
• Butakov N., Vakhrushev A., Khodorchenko M., Zakharova A., Ryzhkov A., Plosskaya O., Gainetdinov A., Damdinov R., Alexandrov D., Nasonov D., Savchenko M. Efficient Deadline-Constrained Scheduling of Distributed AutoML Pipelines ....... 87
• Dementiev Y. I., Ignatiev A. A., Sidelnik V. A., Smal A. V., Ushakov M. S. Advances in the Half-Duplex Communication Complexity ....... 100
• Elistratov S., Podivilov A., Iuzhakov T., Vetrov D. Directional Noise and Implicit Bias in Lion ....... 134
• Galimzianova D., Gorovaia S., Vikhorev D., Yamshchikov I. P., Zhemchuzhina E. CleanComedy: Creating friendly humor through generative techniques ....... 153
• Garipov R., Zabashta A., Filchenkov A. Deep learning architecture for processing tabular data to extract meta-features ....... 178
• Iov I. L., Nikitin N. O. Feature Engineering Pipeline Optimisation in AutoML Workflow using Large Language Models ....... 191
• Laetin A. A., Pulfer B., Smirnov S. K. Hierarchical Initialization of Temporal Scales in Mamba-2 for Multi-Scale Sequence Modeling ....... 223
• Molodetskikh I. A., Malyshev K. V., Mirgaleev M. R., Bogatyrev E. N., Vatolin D. S. Methodology for Collecting and Analyzing Subjective Prominence of Artifacts in Learning-Based Super-Resolution ....... 235
• Nikolenko S. I. Data Filtering and Evaluation Methods for Large Language Models ....... 254}
• Nikolenko S. I. Influence Functions for Data Evaluation in Large Language Models ....... 280
• Oblovatny R., Kuleshova A., Polev K., Zaytsev A. Probabilistic distances-based hallucination detection in LLMs with RAG ....... 327
• Orlov E., Malykh V. Please Upvote: Stack Overflow Answer Ranking in a Text-only Setting ....... 350
• Shumitskaya E. A., Vatolin D. S., Antsiferova A. V. Exploiting stochastic Weaknesses of Randomized Smoothing in Regression Models: Adaptive Attack Analysis and Empirical Limitations ....... 381
• Reviews ....... 398

---

• Paging 411 pp.

---

Back to the Petersburg Department of Steklov Institute of Mathematics