On the Way to Controllable Text Summarization in Russian

This paper is devoted to controllable text summarization in Russian. We explore the effectiveness of a mixture-of-experts architecture for the Russian language. Namely, we adapt the HydraSum method, which involves multiple decoders, for Russian. We compare the performance of this approach with fine-tuning pre-trained language models and show that HydraSum is applicable for text summarization in Russian, yielding good-quality and stylistically diverse summaries.