What properties of reasoning supervision are associated with improved downstream model quality?
M Langner, D Pihulski, JK Eliasz, M Rajkowski, P Kazienko, M Piasecki, J Kocoń, T Ferdinan
ICCS 2026 276–292
2026
Dzmitry Pihulski
Phd Student
Pracownik CLARIN-PL oraz Doktorant Katedry
Research interests
- large language models
- devops
- latent reasoning
- reasoning in llms
Research groups
Education
- 2025
Eng, Applied Mathematics
Wroclaw University of Science and Technology
- 2026
Masters, Trusted AI
Wroclaw University of Science and Technology
Publications2
Breaking the illusion of reasoning in Polish LLMs: quality over quantity of thought
D Pihulski, M Langner, J Eliasz, P Kazienko, J Kocoń, T Ferdinan, …
EACL 2026 1796–1811
2026