FB 6 Mathematik/Informatik/Physik

Institut für Mathematik


Navigation und Suche der Universität Osnabrück


Hauptinhalt

Topinformationen

Study Project: Exploring Physical Understanding in LLMs (Part I)

8.3202

Dozenten

Beschreibung

Overview:
This study project aims to delve into the fascinating debate around whether Large Language Models (LLMs) can develop an understanding of the physical world. We'll explore the potential of multimodal LLMs, like GPT-4, which incorporate both text and visual inputs, to enhance their grounding in language and physical concepts.

Structure:
The project will be structured into three main phases, with weekly meetings focusing on different aspects of learning and research:

Part 1 - Background Learning:
Introduction to LLMs and their capabilities in next-word prediction and multimodal learning. Students will receive learning packages covering the foundational theories and the evolution of LLMs.
Deliverables: Each student will prepare a brief presentation summarizing the key concepts and their implications for understanding physical reality.

Part 2 - Literature Review:
Focus on discussing seminal and recent papers that are pivotal to understanding the current landscape of LLMs’ cognitive abilities, especially in physical understanding. This phase includes identifying gaps in the current research.
Deliverables: In small groups, students will present summaries and critiques of selected papers, highlighting methodologies, findings, and areas needing further exploration.

Part 3 - Research Operationalization and Benchmark Development:
Small groups will operationalize research questions based on identified gaps from the literature review. This phase will involve hands-on work with the GRAP evaluation benchmark as detailed in the paper available at https://arxiv.org/abs/2311.09048
Deliverables: Each group will develop enhancements or new features for the GRAP benchmark, aiming to address specific challenges in evaluating the physical understanding of LLMs. The project concludes with groups presenting their proposed solutions and implementations.

Weitere Angaben

Ort: nicht angegeben
Zeiten:
Erster Termin:
Veranstaltungsart: Studienprojekt (Offizielle Lehrveranstaltungen)

Studienbereiche

  • Cognitive Science > Master-Programm
  • Human Sciences (e.g. Cognitive Science, Psychology)

Past and Forthcoming Events

Publications

  • Asymptotics of a time-bounded cylinder model, with N. Aschenbruck and S. Bussmann, Probability in the Engineering and Informational Sciences, https://doi.org/10.1017/S0269964822000420
  • The method of cumulants for the normal approximation, with S. Jansen and K. Schubert, Probability Surveys 2022, Vol. 19, 185-270, https://doi.org/10.1214/22-PS7
  • Sedentary Random Waypoint, with C. Betken, arXiv:2009.02941
  • The Impact of Bit Errors on Intra-Session Network Coding with Heterogeneous Packet Lengths, with B. Schütz, N. Aschenbruck, S. Bussmann and M. Juhnke-Kubitzke, Proc. of the 45th IEEE LCN Symposium on Emerging Topics in Networking LCN, virtually hosted in Sydney, Australia, Nov. 16–19, 2020.
  • Stationarity for the Small World in Motion Mobility Model, with Nils Aschenbruck, Christian Heiden und Matthias Schwamborn, MSWIM '19: Proceedings of the 22nd International ACM Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems, Nov 25-29, 2019, https://doi.org/10.1145/3345768.3355935
  • Crossing Numbers and Stress of Random Graphs, with Markus Chimani and Matthias Reitzner, In Proceedings 26th International Symposium, GD 2018, Barcelona, Spain, 255--268, 2018 available here and for an extended journal version here: https://arxiv.org/abs/1808.07558
  • Fluctuations in a general preferential attachment model via Stein's method, with Carina Betken and Marcel Ortgiese, Random Structures & algorithms, vol.55, no.4, 2019 available here
  • Connection times in large ad-hoc mobile networks, Bernoulli, vol.22, no.4, 2143--2176, 2016 available here
    with Gabriel Faraud, Wolfgang König
  • The random disc thrower problem, Proceedings of the 90th European Study Group Mathematics with Industry, 59-78, 2013  available here with T. van der Aalst, D. Denteneer, M. Hong Duong, R. J. Kang, M. Keane, J. Kool, I. Kryven, T. Meyfroyt, T. Müller, G. Regts, J. Tomczyk
  • Edge fluctuations of eigenvalues of Wigner matrices, High Dimensional Probability VI: the Banff volume, Progress in Probability, vol.66, 261-275, Springer, Basel, 2013 available here
    with Peter Eichelsbacher
  • Moderate deviations for the determinant of Wigner matrices, Dedicated to Friedrich Götze on the occasion of his sixtieth birthday, Limit Theorems in Probability, Statistics and Number Theory, Springer Proceedings in Mathematics & Statistics, vol.42, 253-275, 2013, available here
    with Peter Eichelsbacher
  • Moderate deviations for the eigenvalue counting function of Wigner matrices, ALEA, Lat. Am. J. Probab. Math. Stat. 10 (1), 27-44, 2013, available here
    with Peter Eichelsbacher
  • Moderate deviations via cumulants, Journal of Theor. Probability, 2012, available here
    with Peter Eichelsbacher
  • Moments of recurrence times for Markov chains, Electronic Comm. Probab., 16(28), 296-303, 2011, available here
    with Frank Aurzada, Marcel Ortgiese, Michael Scheutzow
  • Moderate deviations in a random graph and for the spectrum of Bernoulli random matrices, Electronic Journal of Probability, Vol. 14, Paper no. 92, 2636-2656, 2009, available here
    with Peter Eichelsbacher
  • Perpendicular transport of charged particles in slab turbulence: recovery of diffusion for realistic wave-spectra?, Journal of Physics G: Nuclear and Particle Physics, 35, 025202, 2008
    with Andreas Shalchi
  • Velocity correlation functions of charged test particles, Journal of Physics G: Nuclear and Particle Physics, 34, 859, 2007
    with Andreas Shalchi