Hi! My name is Ilya Gusev.
I am interested in everything around language models and agents. If you need help with any of my projects or you would like to collaborate, do not hesitate to contact me.
How to reach me: @YallenGusev at Telegram.
My Telegram channel: @senior_augur.
Projects:
- holosophos, multi-agent system for writing papers
- codearkt, CodeAct-based agentic framework
- academia_mcp, MCP server with tools for scientific research
- pingpong, benchmark for role-playing LLMs
- saiga_bot, Telegram bot supporting different LLMs
- saiga, training code for Saiga language models
- memetron3000, language models for memes generation
- aika, amateur-level chess engine in C++
- tgcontest, news clustering system in C++
- summarus, models for abstractive and extractive summarization
- rnnmorph, morphological analyzer for Russian
- russ, library for Russian word stress detection
- rupo, library for analysis and generation of poems in Russian
Articles:
- substack, How I Taught AI to Make Memes
- habr, О «раздутом пузыре» нейросетей
- notion, Yet another language model?
- medium, Poems, Flowers, and Dragons at EMNLP 2022
- medium, News Aggregator in 2 weeks
- notion, Beyond OpenAI CLIP
- habr, Цикл про автоматическое реферирование
- habr, Извлекаем суть новости. Опыт Яндекса
- habr, Новостной агрегатор за две недели
- habr, Как научить свою нейросеть анализировать морфологию
- youtube, Как научить нейросеть генерировать стихи
- habr, Как научить свою нейросеть генерировать стихи
Papers:
- PingPong: A Benchmark for Role-Playing Language Models with User Emulation and Multi-Model Evaluation, repo
- Russian Texts Detoxification with Levenshtein Editing, repo
- HeadlineCause: A Dataset of News Headlines for Detecting Causalities, repo
- Russian News Clustering and Headline Selection Shared Task, repo
- Advances of Transformer-Based Models for News Headline Generation
- Dataset for Automatic Summarization of Russian News, repo
- Importance of copying mechanism for news headline generation
- Improving part-of-speech tagging via multi-task learning and character-level word representations
Links:
- Google Scholar: link
- Orchid ID: 0000-0002-8930-729X
- Donate (from Russia): link
- Donate (not from Russia): link