Search

Human learning thrives on the ability to learn from mistakes, adapt through feedback, and refine understanding-processes often missing in static machine learning models. In this work, we introduce Composite Learning Units (CLUs) designed to transform reasoners, such as Large Language Models (LLMs), into learners capable of generalized, continuous learning without conventional parameter updates while enhancing their reasoning abilities through continual interaction and feedback. CLUs are built on an architecture that allows a reasoning model to maintain and evolve a dynamic knowledge repository: a General Knowledge Space for broad, reusable insights and a Prompt-Specific Knowledge Space for task-specific learning. Through goal-driven interactions, CLUs iteratively refine these knowledge spaces, enabling the system to adapt dynamically to complex tasks, extract nuanced insights, and build upon past experiences autonomously. We demonstrate CLUs' effectiveness through a cryptographic reasoning task, where they continuously evolve their understanding through feedback to uncover hidden transformation rules. While conventional models struggle to grasp underlying logic, CLUs excel by engaging in an iterative, goal-oriented process. Specialized components-handling knowledge retrieval, prompt generation, and feedback analysis-work together within a reinforcing feedback loop. This approach allows CLUs to retain the memory of past failures and successes, adapt autonomously, and apply sophisticated reasoning effectively, continually learning from mistakes while also building on breakthroughs.

Iterative human engagement is a common and effective means of leveraging the advanced language processing power of large language models (LLMs). Using well-structured prompts in a conversational manner, human users can effectively influence an LLM to develop more thoughtful and accurate responses. Motivated by this insight, we propose the Iteration of Thought (IoT) framework for enhancing LLM responses by generating "thought"-provoking prompts vis a vis an input query and the current iteration of an LLM's response. Unlike static or semi-static approaches, e.g. Chain of Thought (CoT) or Tree of Thoughts (ToT), IoT adapts its reasoning path dynamically, based on evolving context, and without generating alternate explorative thoughts which are ultimately discarded. The three components of the IoT framework are (1) an Inner Dialogue Agent (IDA) responsible for generating instructive, context-specific prompts; (2) an LLM Agent (LLMA) that processes these prompts to refine its responses; and (3) an iterative prompting loop that implements a conversation between the former two components. We introduce two variants of our framework: Autonomous Iteration of Thought (AIoT), where an LLM decides when to stop iterating, and Guided Iteration of Thought (GIoT), which always forces a fixed number iterations. We investigate the performance of IoT across various datasets, spanning complex reasoning tasks from the GPQA dataset, explorative problem-solving in Game of 24, puzzle solving in Mini Crosswords, and multi-hop question answering from the HotpotQA dataset. Our results show that IoT represents a viable paradigm for autonomous response refinement in LLMs, showcasing significant improvements over CoT and thereby enabling more adaptive and efficient reasoning systems that minimize human intervention.

Deep neural networks (DNNs) offer plenty of challenges in executing efficient computation at edge nodes, primarily due to the huge hardware resource demands. The article proposes HYDRA, hybrid data multiplexing, and runtime layer configurable DNN accelerators to overcome the drawbacks. The work proposes a layer-multiplexed approach, which further reuses a single activation function within the execution of a single layer with improved Fused-Multiply-Accumulate (FMA). The proposed approach works in iterative mode to reuse the same hardware and execute different layers in a configurable fashion. The proposed architectures achieve reductions over 90% of power consumption and resource utilization improvements of state-of-the-art works, with 35.21 TOPSW. The proposed architecture reduces the area overhead (N-1) times required in bandwidth, AF and layer architecture. This work shows HYDRA architecture supports optimal DNN computations while improving performance on resource-constrained edge devices.

We show in this letter how a heavy $(\mathcal{O}(TeV))$ invisible $Z'$ gauge boson that will practically be out of reach of the Large Hadron Collider (LHC), can be discovered at the future muon collider. The new force carrier has a relatively stronger coupling with the beyond standard model (BSM) sector, while its interaction with the SM fields is much weaker. This weaker coupling is induced through mixing mechanisms, specifically via gauge kinetic mixing and the $Z-Z'$ mixing. We consider a scenario where the new gauge boson decays mostly to charge-neutral long-lived particles and/or dark matter (DM). We show how producing and detecting this heavier invisible $Z'$, that will be beyond the reach of even the very high luminosity LHC, becomes possible if it is produced in association with an energetic photon at the future muon collider. The on-shell production of the $Z'$ will lead to a peak in the photon energy distribution, following the so-called radiative return phenomena and can lead to the accurate determination of the $Z'$ mass and its interaction with SM particles., Comment: 6 pages, 4 figures, 5 tables

This paper introduces a Scalable Hierarchical Aware Convolutional Neural Network (SHA-CNN) model architecture for Edge AI applications. The proposed hierarchical CNN model is meticulously crafted to strike a balance between computational efficiency and accuracy, addressing the challenges posed by resource-constrained edge devices. SHA-CNN demonstrates its efficacy by achieving accuracy comparable to state-of-the-art hierarchical models while outperforming baseline models in accuracy metrics. The key innovation lies in the model's hierarchical awareness, enabling it to discern and prioritize relevant features at multiple levels of abstraction. The proposed architecture classifies data in a hierarchical manner, facilitating a nuanced understanding of complex features within the datasets. Moreover, SHA-CNN exhibits a remarkable capacity for scalability, allowing for the seamless incorporation of new classes. This flexibility is particularly advantageous in dynamic environments where the model needs to adapt to evolving datasets and accommodate additional classes without the need for extensive retraining. Testing has been conducted on the PYNQ Z2 FPGA board to validate the proposed model. The results achieved an accuracy of 99.34%, 83.35%, and 63.66% for MNIST, CIFAR-10, and CIFAR-100 datasets, respectively. For CIFAR-100, our proposed architecture performs hierarchical classification with 10% reduced computation while compromising only 0.7% accuracy with the state-of-the-art. The adaptability of SHA-CNN to FPGA architecture underscores its potential for deployment in edge devices, where computational resources are limited. The SHA-CNN framework thus emerges as a promising advancement in the intersection of hierarchical CNNs, scalability, and FPGA-based Edge AI.

Memory management is necessary with the increasing number of multi-connected AI devices and data bandwidth issues. For this purpose, high-speed multi-port memory is used. The traditional multi-port memory solutions are hard-bounded to a fixed number of ports for read or write operations. In this work, we proposed a pseudo-quad-port memory architecture. Here, ports can be configured (1-port, 2-port, 3-port, 4-port) for all possible combinations of read/write operations for the 6T static random access memory (SRAM) memory array, which improves the speed and reduces the bandwidth for data transfer. The proposed architecture improves the bandwidth of data transfer by 4x. The proposed solution provides 1.3x and 2x area efficiency as compared to dual-port 8T and quad-port 12T SRAM. All the design and performance analyses are done using 65nm CMOS technology.

This paper presents the Hybrid Overestimating Approximate Adder designed to enhance the performance in processing engines, specifically focused on edge AI applications. A novel Plus One Adder design is proposed as an incremental adder in the RCA chain, incorporating a Full Adder with an excess 1 alongside inputs A, B, and Cin. The design approximates outputs to 2 bit values to reduce hardware complexity and improve resource efficiency. The Plus One Adder is integrated into a dynamically reconfigurable HOAA, allowing runtime interchangeability between accurate and approximate overestimation modes. The proposed design is demonstrated for multiple applications, such as Twos complement subtraction and Rounding to even, and the Configurable Activation function, which are critical components of the Processing engine. Our approach shows 21 percent improvement in area efficiency and 33 percent reduction in power consumption, compared to state of the art designs with minimal accuracy loss. Thus, the proposed HOAA could be a promising solution for resource-constrained environments, offering ideal trade-offs between hardware efficiency vs computational accuracy.

We study a scotogenic model augmented with an additional U(1) gauge and a discrete Z2 symmetry. The lightest Z2-odd particle in our model becomes the dark matter (DM) candidate while tiny neutrino masses are realized at one loop. We explore the parameter space of the model for which the DM relic density is satisfied, and the correct low-energy neutrino observables are reproduced. The extended gauge symmetry includes beyond Standard Model (SM) particle spectrum consisting of vector-like fermions and scalars. We also highlight possible collider signatures of these particles at the LHC., Comment: 43 pages, 13 figures, 13 tables

We study the prospect of heavy singlet neutrinos as a dark matter (DM) candidate within a neutrinophilic U(1) model, where the Standard Model (SM) is extended with a U(1) gauge symmetry, and neutrino mass and oscillation parameters are explained through an inverse see-saw mechanism. The lightest of the heavy neutrinos plays the role of the DM while the newly introduced scalars and the extra gauge boson Z' act as mediators between the dark sector and the SM sector. We show the range of model parameters where this DM candidate can be accommodated in the Weakly Interacting Massive Particle (WIMP) or Feebly Interacting Massive Particle (FIMP) scenario. The observed DM relic density is achieved via the new gauge boson and singlet scalar portals in the WIMP scenario whereas within the FIMP scenario, these two particles assume a distinct yet pivotal role in generating the observed relic density of dark matter., Comment: 32 pages, 14 figures, 1 table

In this letter, we report the observational constraints on a Bianchi type I anisotropic extension of $w$CDM model with spatial curvature from observational data including Baryon Acoustic Oscillations (BAO), Cosmic chronometers (CC), Big Bang nucleosynthesis (BBN), Pantheon+ (PP) compilation of SNe Ia and SH0ES Cepheid host distance anchors. The anisotropy is found to be of the order $10^{-13}$, which interplay with spatial curvature to reduce $H_0$ tension by $\sim 1\sigma$ as found in the analyses with BAO+CC+BBN+PP combination of data, while no significant effect of anisotropy is observed with BAO+CC+BBN+PPSH0ES combination of data. A closed Universe is favored by $w$CDM as well as anisotropic $w$CDM models with spatial curvature in analyses with BAO+CC+BBN+PP combination of data. An observation of an open Universe from $w$CDM model with spatial curvature in analyses with BAO+CC+BBN+PPSH0ES combination of data and a closed Universe from anisotropic $w$CDM model with curvature in analyses with same combination of data is made. The quintessence form of dark energy is favored at 95\% CL in both analyses., Comment: 7 pages, 4 figures, 1 table

The increasing demand for computational photography and imaging on mobile platforms has led to the widespread development and integration of advanced image sensors with novel algorithms in camera systems. However, the scarcity of high-quality data for research and the rare opportunity for in-depth exchange of views from industry and academia constrain the development of mobile intelligent photography and imaging (MIPI). Building on the achievements of the previous MIPI Workshops held at ECCV 2022 and CVPR 2023, we introduce our third MIPI challenge including three tracks focusing on novel image sensors and imaging algorithms. In this paper, we summarize and review the Nighttime Flare Removal track on MIPI 2024. In total, 170 participants were successfully registered, and 14 teams submitted results in the final testing phase. The developed solutions in this challenge achieved state-of-the-art performance on Nighttime Flare Removal. More details of this challenge and the link to the dataset can be found at https://mipi-challenge.org/MIPI2024/., Comment: CVPR 2024 Mobile Intelligent Photography and Imaging (MIPI) Workshop--Nighttime Flare Removal Challenge Report. Website: https://mipi-challenge.org/MIPI2024/

(111)NiO epitaxial films are grown on c-sapphire substrates at various growth temperatures ranging from room-temperature to 600C using pulsed laser deposition (PLD) technique. Two series of samples, where different laser fluences are used to ablate the target, are studied here. Films grown with higher laser fluence, are found to be embedded with Ni-clusters crystallographically aligned with the (111)NiO matrix. While the layers grown with lower laser energy density exhibit p-type conductivity specially at low growth temperatures. X-ray diffraction study shows the coexistence of biaxial compressive and tensile hydrostatic strains in these samples, which results in an expansion of the lattice primarily along the growth direction. This effective uniaxial expansion {epsilon}_perpendicular increases with the reduction of the growth temperature. Band gap of these samples is found to decrease linearly with {epsilon}_perpendicular. This result is validated by density functional theory (DFT) calculations. Experimental findings and the theoretical study further indicate that V_Ni + O_I and V_O + Ni_I complexes exist as the dominant native defects in samples grown with Ni-deficient (low laser fluence) and Ni-rich (high laser fluence) conditions, respectively. P-type conductivity observed in the samples grown in Ni-deficient condition is more likely to be resulting from V_Ni + O_I defects than Ni-vacancies (V_Ni)., Comment: 9 pages and 9 figures (Main manuscript), 4 pages and 4 figures (supplemental material)

In this paper, we introduce the cumulative past information generating function (CPIG) and relative cumulative past information generating function (RCPIG). We study its properties. We establish its relation with generalized cumulative past entropy (GCPE). We defined CPIG stochastic order and its relation with dispersive order. We provide the results for the CPIG measure of the convoluted random variables in terms of the measures of its components. We found some inequality relating to Shannon entropy, CPIG and GCPE. Some characterization and estimation results are also discussed regarding CPIG. We defined divergence measures between two random variables, Jensen-cumulative past information generating function(JCPIG), Jensen fractional cumulative past entropy measure, cumulative past Taneja entropy, and Jensen cumulative past Taneja entropy information measure.

The rise of cloud computing has spurred a trend of transferring data storage and computational tasks to the cloud. To protect confidential information such as customer data and business details, it is essential to encrypt this sensitive data before cloud storage. Implementing encryption can prevent unauthorized access, data breaches, and the resultant financial loss, reputation damage, and legal issues. Moreover, to facilitate the execution of data mining algorithms on the cloud-stored data, the encryption needs to be compatible with domain computation. The $k$-nearest neighbor ($k$-NN) computation for a specific query vector is widely used in fields like location-based services. Sanyashi et al. (ICISS 2023) proposed an encryption scheme to facilitate privacy-preserving $k$-NN computation on the cloud by utilizing Asymmetric Scalar-Product-Preserving Encryption (ASPE). In this work, we identify a significant vulnerability in the aforementioned encryption scheme of Sanyashi et al. Specifically, we give an efficient algorithm and also empirically demonstrate that their encryption scheme is vulnerable to the ciphertext-only attack (COA).

The electronic band structure of V$_2$O$_5$ is calculated using an all-electron quasiparticle self-consistent (QS) $GW$ method, including electron-hole ladder diagrams in the screening of $W$. The optical dielectric function calculated with the Bethe-Salpeter equation exhibits excitons with large binding energy, consistent with spectroscopic ellipsometry data and other recent calculations. Sharp peaks in the direction perpendicular to the layers at high energy are found to be an artifact of the truncation of the numbers of bands included in the BSE calculation of the macroscopic dielectric function. The $\varepsilon_1(\omega=0)$ gives indices of refraction in good agreement with experiment. The excitons are charge transfer excitons with the hole primarily on oxygen and electrons on vanadium, but depending on which exciton, the distribution over different oxygens changes. The exciton wave functions have a spread of about 5-15\AA, with asymmetric character for the electron distribution around the hole depending on which oxygen the hole is fixed at. The monolayer quasiparticle gap increases inversely proportional to interlayer distance once the initial interlayer covalent couplings are removed which is thanks to the long-range nature of the self-energy and the reduced screening in a 2D system. The optical gap on the other hand is relatively independent of interlayer spacing because of the compensation between the self-energy gap shift and the exciton binding energy, both of which are proportional to the screened Coulomb interaction $\hat{W}$. Recent experimental results on very thin layer V$_2$O$_5$ obtained by chemical exfoliation provide experimental support for an increase in gap., Comment: 20 pages, 16 figures, data files available on https://github.com/Electronic-Structure-Group/v2o5-quasiparticle-exciton

In this paper, we derive observational constraints on an anisotropic $w$CDM model from observational data including Baryonic Acoustic Oscillations (BAOs), Cosmic Chronometer (CC), Big Bang Nucleosynthesis (BBN), Pantheon Plus (PP) compilation of Type Ia supernovae, and SH0ES Cepheid host distance anchors. We find that anisotropy is of the order $10^{-13}$, and its presence in the $w$CDM model reduces $H_0$ tension by $\sim 2\sigma$ and $\sim 1\sigma$ in the analyses with BAO+CC+BBN+PP and BAO+CC+BBN+PPSH0ES data combinations, respectively. In both analyses, the quintessence form of dark energy is favored at 95\% CL., Comment: 15 pages, 6 figures, 4 tables. Minor corrections and a subsection of statistical model comparison is added. Matches with the published version of journal

KCoO2 has been found in 1975 to exist in a unique structure with P4/nmm spacegroup with Co in a square pyramidal coordination with the Co atoms in the plane linked by O in a square arrangement reminiscent of the cuprates but its electronic structure has not been studied until now. Unlike Co atoms in LiCoO2 and NaCoO2 in octahedral coordination, which are non-magnetic band structure insulators, the unusual coordination of d6 Co^{3+} in KCoO2 is here shown to lead to a magnetic stabilization of an insulating structure with high magnetic moments of 4{\mu}B per Co. The electronic band structure is calculated using the quasiparticle self-consistent (QS)GW method and the basic formation of magnetic moments is explained in terms of the orbital decomposition of the bands. The optical dielectric function is calculated using the Bethe-Salpeter equation including only transitions between equal spin bands. The magnetic moments are shown to prefer an antiferromagnetic ordering along the [110] direction. Exchange interactions are calculated from the transverse spin susceptibility and a rigid spin approximation. The N\'eel temperature is estimated using the mean-field and Tyablikov methods and found to be between approximately 100 and 250 K. The band structure in the AFM ordering can be related to the FM ordering by band folding effects. The optical spectra are similar in both structures and show evidence of excitonic features below the quasiparticle gap of about 4 eV., Comment: 11 pages, 13 figures

We study the problem of selecting a user equipment (UE) and a beam for each access point (AP) for concurrent transmissions in a millimeter wave (mmWave) network, such that the sum of weighted rates of UEs is maximized. We prove that this problem is NP-complete. We propose two algorithms -- Markov Chain Monte Carlo (MCMC) based and local interaction game (LIG) based UE and beam selection -- and prove that both of them asymptotically achieve the optimal solution. Also, we propose two fast greedy algorithms -- NGUB1 and NGUB2 -- for UE and beam selection. Through extensive simulations, we show that our proposed greedy algorithms outperform the most relevant algorithms proposed in prior work and perform close to the asymptotically optimal algorithms., Comment: 16 pages, 3 figures

In many statistical studies, the measure of uncertainties like entropy, extropy, varentropy and varextropy of a distribution function is of prime interest. This paper proposes estimators of extropy and varextropy. Proposed estimators are consistent. Based on extropy estimator, a test of symmetry is given. The proposed test has the advantage that we do not need to estimate the centre of symmetry. The critical value and power of the proposed test statistics have been obtained. The test procedure has been implemented on six real-life data sets to verify its performance in identifying the symmetric nature., Comment: arXiv admin note: text overlap with arXiv:2209.06703