Search

With the growing popularity of generative AI for images, video, and music, we witnessed models rapidly improve in quality and performance. However, not much attention is paid towards enabling AI's ability to "be creative". In this study, we implemented and simulated the systems model of creativity (proposed by Csikszentmihalyi) using virtual agents utilizing large language models (LLMs) and text prompts. For comparison, the simulations were conducted with the "virtual artists" being: 1)isolated and 2)placed in a multi-agent system. Both scenarios were compared by analyzing the variations and overall "creativity" in the generated artifacts (measured via a user study and LLM). Our results suggest that the generative agents may perform better in the framework of the systems model of creativity.

We proved that the boundary Dehn twist on the Milnor fiber $M_c(2, q, r)$ is an exotic diffeomorphism relative to the boundary if $q, r$ are odd, coprime integers bigger than $3$ and $(q-1)(r-1)/4$ is an odd number. The proof is given by comparing the family relative Bauer--Furuta invariants of the mapping torus., Comment: 12 pages

Consider a one-dimensional diffusion process which has state-dependent drift and deviation and is reflected at the origin, which is called a one-side reflected diffusion or simply reflected diffusion. We are particularly interested in the case that its drift and deviation are discontinuous. We define this reflected diffusion as the solution of a stochastic integral equation, and find conditions for its positive recurrence, We then derive its stationary distribution under these conditions. As a related problem, we also consider the case that it is reflected at two sides, which is called a two-sides reflected diffusion. Its existence, positive recurrence and stationary distribution are similarly studied. In the literature, these problems are studied through a state-dependent diffusion on the whole line particularly when the drift and deviation are discontinuous. However, the reflected process itself is not defined in such a study. Thus, the stationary distribution has not been fully studied for a general state-dependent reflected diffusion. We aim to fills this insufficiency and to make the stationary distributions of reflected diffusions widely available in application., Comment: Submitted for publication

We propose ratio divergence (RD) learning for discrete energy-based models, a method that utilizes both training data and a tractable target energy function. We apply RD learning to restricted Boltzmann machines (RBMs), which are a minimal model that satisfies the universal approximation theorem for discrete distributions. RD learning combines the strength of both forward and reverse Kullback-Leibler divergence (KLD) learning, effectively addressing the "notorious" issues of underfitting with the forward KLD and mode-collapse with the reverse KLD. Since the summation of forward and reverse KLD seems to be sufficient to combine the strength of both approaches, we include this learning method as a direct baseline in numerical experiments to evaluate its effectiveness. Numerical experiments demonstrate that RD learning significantly outperforms other learning methods in terms of energy function fitting, mode-covering, and learning stability across various discrete energy-based models. Moreover, the performance gaps between RD learning and the other learning methods become more pronounced as the dimensions of target models increase., Comment: 14 pages, 19 figures

There are several challenges in developing a model for multi-tasking humanoid control. Reinforcement learning and imitation learning approaches are quite popular in this domain. However, there is a trade-off between the two. Reinforcement learning is not the best option for training a humanoid to perform multiple behaviors due to training time and model size, and imitation learning using kinematics data alone is not appropriate to realize the actual physics of the motion. Training models to perform multiple complex tasks take long training time due to high DoF and complexities of the movements. Although training models offline would be beneficial, another issue is the size of the dataset, usually being quite large to encapsulate multiple movements. There are few implementations of transformer-based models to control humanoid characters and predict their motion based on a large dataset of recorded/reference motion. In this paper, we train a GPT on a large dataset of noisy expert policy rollout observations from a humanoid motion dataset as a pre-trained model and fine tune that model on a smaller dataset of noisy expert policy rollout observations and actions to autoregressively generate physically plausible motion trajectories. We show that it is possible to train a GPT-based foundation model on a smaller dataset in shorter training time to control a humanoid in a realistic physics environment to perform human-like movements.

We investigate approximation algorithms for several fundamental optimization problems on geometric packing. The geometric objects considered are very generic, namely $d$-dimensional convex fat objects. Our main contribution is a versatile framework for designing efficient approximation schemes for classic geometric packing problems. The framework effectively addresses problems such as the multiple knapsack, bin packing, multiple strip packing, and multiple minimum container problems. Furthermore, the framework handles additional problem features, including item multiplicity, item rotation, and additional constraints on the items commonly encountered in packing contexts. The core of our framework lies in formulating the problems as integer programs with a nearly decomposable structure. This approach enables us to obtain well-behaved fractional solutions, which can then be efficiently rounded. By modeling the problems in this manner, our framework offers significant flexibility, allowing it to address a wide range of problems and incorporate additional features. To the best of our knowledge, prior to this work, the known results on approximation algorithms for packing problems were either highly fixed for one problem or restricted to one class of objects, mainly polygons and hypercubes. In this sense, our framework is the first result with a general toolbox flavor in the context of approximation algorithms for geometric packing problems. Thus, we believe that our technique is of independent interest, being possible to inspire further work on geometric packing.

In the classic realm of impact mitigation, targeting different impact scenarios with a universally designed device still remains an unassailable challenge. In this study, we delve into the untapped potential of Resch-patterned origami for impact mitigation, specifically considering the adaptively reconfigurable nature of the Resch origami structure. Our unit-cell-level analyses reveal two distinctive modes of deformation, each characterized by contrasting mechanical responses: the folding mode that displays monostability coupled with strain-hardening, and the unfolding mode that manifests bistability, facilitating energy absorption through snap-through dynamics. Drop tests further unveil a novel dynamic bifurcation phenomenon, where the origami switches between folding and unfolding depending on impact speed, thereby showcasing its innate self-reconfigurability in a wide range of dynamic events. The tessellated meter-scale Resch structure mimicking an automotive bumper inherits this dynamically bifurcating behavior, demonstrating the instantaneous morphing into favorable deformation mode to minimize the peak acceleration upon impact. This suggests a self-adaptive and universally applicable impact-absorbing nature of the Resch-patterned origami system. We believe that our findings pave the way for developing smart, origami-inspired impact mitigation devices capable of real-time response and adaptation to external stimuli, offering insights into designing universally protective structures with enhanced performance in response to various impact scenarios.

In Braverman et al. (2024), the authors prove that the stationary distribution of a multiclass queueing network converges to the stationary distribution of a semimartingale reflecting Brownian motion (SRBM) in heavy traffic. Among the sufficient conditions for the convergence is that the reflection matrix $R$ of the SRBM is "tight". In this paper, we study how we can verify this tightness of $R$. For a $2$-dimensional SRBM, we give necessary and sufficient conditions for $R$ to be tight, while, for a general dimension, we only give sufficient conditions. We then apply these results to the SRBMs arising from the diffusion approximations of multiclass queueing networks with static buffer priority service disciplines that are studied in Braverman et al. (2024). It is shown that $R$ is always tight for this network with two stations if $R$ is defined and completely-S. For the case of more than two stations, it is shown that $R$ is tight for reentrant lines with last-buffer-first-service (LBFS) discipline, but it is not always tight for reentrant line with first-buffer-first-service (FBFS) discipline.

How are emotions formed? Through extensive debate and the promulgation of diverse theories , the theory of constructed emotion has become prevalent in recent research on emotions. According to this theory, an emotion concept refers to a category formed by interoceptive and exteroceptive information associated with a specific emotion. An emotion concept stores past experiences as knowledge and can predict unobserved information from acquired information. Therefore, in this study, we attempted to model the formation of emotion concepts using a constructionist approach from the perspective of the constructed emotion theory. Particularly, we constructed a model using multilayered multimodal latent Dirichlet allocation , which is a probabilistic generative model. We then trained the model for each subject using vision, physiology, and word information obtained from multiple people who experienced different visual emotion-evoking stimuli. To evaluate the model, we verified whether the formed categories matched human subjectivity and determined whether unobserved information could be predicted via categories. The verification results exceeded chance level, suggesting that emotion concept formation can be explained by the proposed model., Comment: This work has been submitted to the IEEE for possible publication. We would like to thank Professor Takayuki Nagai for useful discussions

XL-Calibur is a balloon-borne Compton polarimeter for X-rays in the $\sim$15-80 keV range. Using an X-ray mirror with a 12 m focal length for collecting photons onto a beryllium scattering rod surrounded by CZT detectors, a minimum-detectable polarization as low as $\sim$3% is expected during a 24-hour on-target observation of a 1 Crab source at 45$^{\circ}$ elevation. Systematic effects alter the reconstructed polarization as the mirror focal spot moves across the beryllium scatterer, due to pointing offsets, mechanical misalignment or deformation of the carbon-fiber truss supporting the mirror and the polarimeter. Unaddressed, this can give rise to a spurious polarization signal for an unpolarized flux, or a change in reconstructed polarization fraction and angle for a polarized flux. Using bench-marked Monte-Carlo simulations and an accurate mirror point-spread function characterized at synchrotron beam-lines, systematic effects are quantified, and mitigation strategies discussed. By recalculating the scattering site for a shifted beam, systematic errors can be reduced from several tens of percent to the few-percent level for any shift within the scattering element. The treatment of these systematic effects will be important for any polarimetric instrument where a focused X-ray beam is impinging on a scattering element surrounded by counting detectors., Comment: Submitted to Astroparticle Physics

We investigate edge and bulk states in Weyl-orbit based quantum Hall effect by measuring a Corbino-type device fabricated from a topological Dirac semimetal (Cd1-xZnx)3As2 film. Clear quantum Hall plateaus are observed when measuring one-sided terminals of the Corbino-type device. This indicates that edge states of the Weyl-orbit quantum Hall effect form closed trajectories consisting of Fermi arcs and chiral zero modes independently on inner and outer sides. On the other hand, the bulk resistance does not diverge at fields where the quantum Hall plateau appears, suggesting that the Weyl orbits in the bulk region are not completely localized when applying electric current through the bulk region.

We consider a single server queue which has a threshold to change its arrival process and service speed by its queue length, which is referred to as a two-level single server queue. This model is motivated by an energy saving problem for a single server queue whose arrival process and service speed are controlled. To get its performance in tractable form, we study the limit of the stationary distribution of the queue length in this two-level queue under scaling in heavy traffic. Except for a special case, this limit corresponds to its diffusion approximation. It is shown that this limiting distribution is truncated exponential (or uniform if the drift is null) below the threshold level and exponential above it under suitably chosen system parameters and generally distributed inter-arrival times and workloads brought by customers. This result is proved under a mild limitation on arrival parameters using the so called BAR approach studied in Braverman, Dai and Miyazawa (2017, 2023) and Miyazawa (2017, 2023). We also intuitively discuss about a diffusion process corresponding to the limit of the stationary distribution under scaling., Comment: 40 pages, 2 figure, submitted for publication. This paper will appear in Advances in Applied Probability 57.4 (December 2025)

We give an infinite family of embeddings of $\mathbb{R} P^2$ to $S^4$ such that they are mutually topologically isotopic however are not smoothly isotopic to each other. Moreover, they are topologically isotopic to the standard $P^2$-knot. To prove that these $P^2$-knots are not smoothly isotopic to each other, we construct a gauge theoretic invariant of embedded surfaces in $4$-manifolds using a variant of the Seiberg--Witten theory, which is called the Real Seiberg--Witten theory., Comment: 45 pages