-
The Good, the Bad, and the Hulk-like GPT: Analyzing Emotional Decisions of Large Language Models in Cooperation and Bargaining Games
Authors:
Mikhail Mozikov,
Nikita Severin,
Valeria Bodishtianu,
Maria Glushanina,
Mikhail Baklashkin,
Andrey V. Savchenko,
Ilya Makarov
Abstract:
Behavior study experiments are an important part of society modeling and understanding human interactions. In practice, many behavioral experiments encounter challenges related to internal and external validity, reproducibility, and social bias due to the complexity of social interactions and cooperation in human user studies. Recent advances in Large Language Models (LLMs) have provided researche…
▽ More
Behavior study experiments are an important part of society modeling and understanding human interactions. In practice, many behavioral experiments encounter challenges related to internal and external validity, reproducibility, and social bias due to the complexity of social interactions and cooperation in human user studies. Recent advances in Large Language Models (LLMs) have provided researchers with a new promising tool for the simulation of human behavior. However, existing LLM-based simulations operate under the unproven hypothesis that LLM agents behave similarly to humans as well as ignore a crucial factor in human decision-making: emotions.
In this paper, we introduce a novel methodology and the framework to study both, the decision-making of LLMs and their alignment with human behavior under emotional states. Experiments with GPT-3.5 and GPT-4 on four games from two different classes of behavioral game theory showed that emotions profoundly impact the performance of LLMs, leading to the development of more optimal strategies. While there is a strong alignment between the behavioral responses of GPT-3.5 and human participants, particularly evident in bargaining games, GPT-4 exhibits consistent behavior, ignoring induced emotions for rationality decisions. Surprisingly, emotional prompting, particularly with `anger' emotion, can disrupt the "superhuman" alignment of GPT-4, resembling human emotional responses.
△ Less
Submitted 5 June, 2024;
originally announced June 2024.
-
Collective States in Molecular Monolayers on 2D Materials
Authors:
Sabrina Juergensen,
Moritz Kessens,
Charlotte Berrezueta-Palacios,
Nikolai Severin,
Sumaya Ifland,
Jürgen P. Rabe,
Niclas S. Mueller,
Stephanie Reich
Abstract:
Collective excited states form in organic two-dimensional layers through the Coulomb coupling of the molecular transition dipole moments. They manifest as characteristic strong and narrow peaks in the excitation and emission spectra that are shifted to lower energies compared to the monomer transition. We study experimentally and theoretically how robust the collective states are against homogeneo…
▽ More
Collective excited states form in organic two-dimensional layers through the Coulomb coupling of the molecular transition dipole moments. They manifest as characteristic strong and narrow peaks in the excitation and emission spectra that are shifted to lower energies compared to the monomer transition. We study experimentally and theoretically how robust the collective states are against homogeneous and inhomogeneous broadening as well as spatial disorder that occur in real molecular monolayers. Using a microscopic model for a two-dimensional dipole lattice in real space we calculate the properties of collective states and their extinction spectra. We find that the collective states persist even for 1-10% random variation in the molecular position and in the transition frequency, with similar peak position and integrated intensity as for the perfectly ordered system. We measure the optical response of a monolayer of the perylene-derivative MePTCDI on two-dimensional materials. On the wide band-gap insulator hexagonal boron nitride it shows strong emission from the collective state with a line width that is dominated by the inhomogeneous broadening of the molecular state. When using the semimetal graphene as a substrate, however, the luminescence is completely quenched. By combining optical absorption, luminescence, and multi-wavelength Raman scattering we verify that the MePTCDI molecules form very similar collective monolayer states on hexagonal boron nitride and graphene substrates, but on graphene the line width is dominated by non-radiative excitation transfer from the molecules to the substrate. Our study highlights the transition from the localized molecular state of the monomer to a delocalized collective state in the two-dimensional molecular lattice that is entirely based on Coulomb coupling between optically active excitations of the electrons or molecular vibrations.
△ Less
Submitted 14 August, 2023; v1 submitted 18 June, 2023;
originally announced June 2023.
-
Temporal Graph Network Embedding with Causal Anonymous Walks Representations
Authors:
Ilya Makarov,
Andrey Savchenko,
Arseny Korovko,
Leonid Sherstyuk,
Nikita Severin,
Aleksandr Mikheev,
Dmitrii Babaev
Abstract:
Many tasks in graph machine learning, such as link prediction and node classification, are typically solved by using representation learning, in which each node or edge in the network is encoded via an embedding. Though there exists a lot of network embeddings for static graphs, the task becomes much more complicated when the dynamic (i.e. temporal) network is analyzed. In this paper, we propose a…
▽ More
Many tasks in graph machine learning, such as link prediction and node classification, are typically solved by using representation learning, in which each node or edge in the network is encoded via an embedding. Though there exists a lot of network embeddings for static graphs, the task becomes much more complicated when the dynamic (i.e. temporal) network is analyzed. In this paper, we propose a novel approach for dynamic network representation learning based on Temporal Graph Network by using a highly custom message generating function by extracting Causal Anonymous Walks. For evaluation, we provide a benchmark pipeline for the evaluation of temporal network embeddings. This work provides the first comprehensive comparison framework for temporal network representation learning in every available setting for graph machine learning problems involving node classification and link prediction. The proposed model outperforms state-of-the-art baseline models. The work also justifies the difference between them based on evaluation in various transductive/inductive edge/node classification tasks. In addition, we show the applicability and superior performance of our model in the real-world downstream graph machine learning task provided by one of the top European banks, involving credit scoring based on transaction data.
△ Less
Submitted 24 August, 2021; v1 submitted 19 August, 2021;
originally announced August 2021.
-
Extrinsic localized excitons in patterned 2D semiconductors
Authors:
D Yagodkin,
K Greben,
A Eljarrat,
S Kovalchuk,
M Ghorbani-Asl,
M Jain,
S Kretschmer,
N Severin,
J P Rabe,
A V Krasheninnikov,
C T Koch,
K I Bolotin
Abstract:
We demonstrate a new localized excitonic state in patterned monolayer 2D semiconductors. This state is not associated with lattice disorder but is extrinsic, i.e. results from external molecules on the material surface. The signature of an exciton associated with that state is observed in the photoluminescence spectrum after electron beam exposure of several 2D semiconductors. The localized state,…
▽ More
We demonstrate a new localized excitonic state in patterned monolayer 2D semiconductors. This state is not associated with lattice disorder but is extrinsic, i.e. results from external molecules on the material surface. The signature of an exciton associated with that state is observed in the photoluminescence spectrum after electron beam exposure of several 2D semiconductors. The localized state, which is distinguished by non-linear power dependence, survives up to room temperature and is patternable down to 20 nm resolution. We probe the response of the new exciton to the changes of electron energy, nanomechanical cleaning, and encapsulation via multiple microscopic, spectroscopic, and computational techniques. All these approaches suggest that the state does not originate from irradiation-induced structural defects or spatially non-uniform strain, as commonly assumed. Instead, we show that it is extrinsic, likely a charge transfer exciton associated with the organic substance deposited onto the 2D semiconductor. By demonstrating that structural defects are not required for the formation of localized excitons, our work opens new possibilities for further understanding of these states and using them for example in chemical sensing and quantum technologies.
△ Less
Submitted 29 March, 2022; v1 submitted 28 May, 2021;
originally announced May 2021.
-
Microstructure and Elastic Constants of Transition Metal Dichalcogenide Monolayers from Friction and Shear Force Microscopy
Authors:
Xiaomin Xu,
Thorsten Schultz,
Ziyu Qin,
Nikolai Severin,
Benedikt Haas,
Sumin Shen,
Jan N. Kirchhof,
Andreas Opitz,
Christoph T. Koch,
Kirill Bolotin,
Jürgen P. Rabe,
Goki Eda,
Norbert Koch
Abstract:
Optical and electrical properties of two-dimensional transition metal dichalcogenides (TMDCs) grown by chemical vapor deposition (CVD) are strongly determined by their microstructure. Consequently, the visualization of spatial structural variations is of paramount importance for future applications. Here we demonstrate how grain boundaries, crystal orientation, and strain fields can unambiguously…
▽ More
Optical and electrical properties of two-dimensional transition metal dichalcogenides (TMDCs) grown by chemical vapor deposition (CVD) are strongly determined by their microstructure. Consequently, the visualization of spatial structural variations is of paramount importance for future applications. Here we demonstrate how grain boundaries, crystal orientation, and strain fields can unambiguously be identified with combined lateral force microscopy (LFM) and transverse shear microscopy (TSM) for CVD-grown tungsten disulfide (WS2) monolayers, on length scales that are relevant for optoelectronic applications. Further, angle-dependent TSM measurements enable us to acquire the fourth-order elastic constants of monolayer WS2 experimentally. Our results facilitate high-throughput and nondestructive microstructure visualization of monolayer TMDCs, insights into their elastic properties, thus providing an accessible tool to support the development of advanced optoelectronic devices based on such two-dimensional semiconductors.
△ Less
Submitted 21 August, 2019;
originally announced August 2019.
-
Pulsed thermal deposition of binary and ternary transition metal dichalcogenide monolayers and heterostructures
Authors:
Niklas Mutz,
Tino Meisel,
Holm Kirmse,
Soohyung Park,
Nikolai Severin,
Jürgen P. Rabe,
Emil List-Kratochvil,
Norbert Koch,
Christoph Koch,
Sylke Blumstengel,
Sergey Sadofev
Abstract:
Application of transition metal dichalcogenides (TMDC) in photonic, optoelectronic or valleytronic devices requires the growth of continuous monolayers, heterostructures and alloys of different materials in a single process. We present a facile pulsed thermal deposition method which provides precise control over layer thickness and stoichiometry of two-dimensional systems. The versatility of the m…
▽ More
Application of transition metal dichalcogenides (TMDC) in photonic, optoelectronic or valleytronic devices requires the growth of continuous monolayers, heterostructures and alloys of different materials in a single process. We present a facile pulsed thermal deposition method which provides precise control over layer thickness and stoichiometry of two-dimensional systems. The versatility of the method is demonstrated on ternary monolayers of Mo$_{1-x}$W$_{x}$S$_{2}$ and on heterostructures combining metallic TaS$_{2}$ and semiconducting MoS$_{2}$ layers. The fabricated ternary monolayers cover the entire composition range of $x$ = 0...1 without phase separation. Band gap engineering and control over the spin-orbit coupling strength is demonstrated by absorption and photoluminescence spectroscopy. Vertical heterostructures are grown without intermixing. The formation of clean and atomically abrupt interfaces is evidenced by high-resolution transmission electron microscopy. Since both the metal components as well as the chalcogenides are thermally evaporated complex alloys and heterostructures can thus be prepared.
△ Less
Submitted 11 February, 2019; v1 submitted 19 November, 2018;
originally announced November 2018.
-
Reconstructing Interaction Potentials in Thin Films from Real-Space Images
Authors:
Jonas Gienger,
Nikolai Severin,
Jürgen P. Rabe,
Igor M. Sokolov
Abstract:
We demonstrate that an inverse Monte Carlo approach allows to reconstruct effective interaction potentials from real-space images. The method is exemplified on monomolecular ethanol-water films imaged with scanning force microscopy (SFM), which provides the spatial distribution of the molecules. Direct Monte Carlo simulations with the reconstructed potential allow for obtaining characteristics of…
▽ More
We demonstrate that an inverse Monte Carlo approach allows to reconstruct effective interaction potentials from real-space images. The method is exemplified on monomolecular ethanol-water films imaged with scanning force microscopy (SFM), which provides the spatial distribution of the molecules. Direct Monte Carlo simulations with the reconstructed potential allow for obtaining characteristics of the system which are unavailable in the experiment, such as the heat capacity of the monomolecularly thin film, and for a prediction of the critical temperature of the demixing transition.
△ Less
Submitted 7 March, 2016; v1 submitted 12 May, 2015;
originally announced May 2015.
-
Evidence for graphene plasmons in the visible spectral range probed by molecules
Authors:
Philipp Lange,
Günter Kewes,
Nikolai Severin,
Oliver Benson,
Jürgen P. Rabe
Abstract:
Graphene is considered to be plasmon active only up to the infrared based on combined tight binding model and random phase approximation calculations. Here we show that the optical properties of graphene as measured by ellipsometry and simulated by density functional theory imply the existence of strongly localized graphene plasmons in the visible with a line width of 0.1 eV. Using small emitters…
▽ More
Graphene is considered to be plasmon active only up to the infrared based on combined tight binding model and random phase approximation calculations. Here we show that the optical properties of graphene as measured by ellipsometry and simulated by density functional theory imply the existence of strongly localized graphene plasmons in the visible with a line width of 0.1 eV. Using small emitters that provide the high wavevectors necessary to excite graphene plasmons at optical frequencies we demonstrate graphene plasmon induced excitation enhancement by nearly 3 orders of magnitude.
△ Less
Submitted 25 April, 2014;
originally announced April 2014.