General Seminar
current | 2026 | 2025 | 2024 | 2023 | 2022 | 2021 | 2020 | 2019 | 2018 | 2017
General Seminar takes place every Monday at 11.15 in Piwnice, radioastronomy seminar room.
Remote participation via the BigBlueButton (BBB).
25 May 2026
“Do LLMs Reason?”
prof. Ulrike Hahn
Birkbeck Univ London
Abstract:
Few debates have been as polarised as the debate about “reasoning” in LLMs. The extent of the disagreement seems noteworthy in and of itself and in need of explanation. The talk details recent work (Hahn, 2026) seeking to clarify underlying terminological and conceptual disagreements that give rise to these seemingly irreconcilable positions.
Preprint: Hahn, 2026, https://zenodo.org/records/18231172
1 June 2026
“Position of X-ray low surface brightness clusters in the cosmic filament network: dependence on filament catalog selection”
dr hab. Mariusz Tarnopolski, prof. UMK
Institute of Astronomy, Nicolaus Copernicus University
Abstract:
Zarattini et al. [Astronomy and Astrophysics, 694 (2025), A256] recently reported that core-excised X-ray surface brightness, SB_X, of low surface brightness clusters (LSBCs) exhibit an anti-correlation with their distance to nearest galactic filament, D_fila. We determine the position of LSBCs with respect to the cosmic network by employing three different filament catalogs. Utilizing the X-ray Unbiased Cluster Sample (XUCS), we investigated how D_fila depends on SB_X for the three filament catalogs. The XUCS samples were relatively small, ranging from 17 to 29 LSBCs. We took into account the uncertainties in SB_X and performed Monte Carlo simulations to assess the probability of a particular pattern being a chance occurrence. We obtain a positive correlation, anti-correlation, and no correlation depending on the filament catalog utilized. Using a permutation test we estimate that, given the sample sizes, there is a 40%-60% probability that the observed patterns are a chance occurrence. By varying the measured values within their standard errors, we find that there is a further 6%-8% probability that the patterns present in the data are spurious. The relation between D_fila and SB_X depends on the filament catalog used: there can be a positive correlation, anti-correlation, or no correlation. This sensitivity indicates that systematic differences in filament finding algorithms can significantly affect environmental metrics such as D_fila. All in all, the current samples are too small to infer the character of the relation between SB_X and D_fila.
8 June 2026
“On the similarity of cosmic structures”
prof. Matthias Bartelmann
Institut für Theoretische Physik, Universität Heidelberg
Abstract:
The Universe is filled with structures of vastly different scales: galaxies themselves span many orders of magnitude in mass and size; they agglomerate in galaxy groups and clusters; and they align along filaments tens of millions of lightyears long. Notwithstanding their vastly different sizes, these gravitationally bound structures are internally very similar: for example, the radial profiles of their dominant, dark-matter density differ by scale, but not by shape. Why is this so, and what can we learn from this kind of universality? Does it depend on the nature of dark matter, details of the cosmological model, or of gravity theory? The talk will discuss approaches to these questions, present partial answers, and lead to conjectures about cosmic structures that remain to be tested.