arXiv subjects

Arnav Das

Publications and source records attributed to Arnav Das.

13 recordsLinked to original sources

arXiv2026

Coresets Before Score Sets: Evaluation-Unsupervised Prompt Subset Selection for LLM Benchmarks

We study LLM benchmark coreset selection: selecting a small subset of prompts over multiple benchmarks whose induced model scores and rankings approximate those obtained from the full benchmark suite. In evaluation-unsupervised benchmark coreset selection (our approach), the selection algorithm uses no model evaluation outcomes, and operates on a fine granularity by producing subsets of prompts over multiple benchmarks rather than producing a sub-collection of entire benchmarks. We use submodular subset selection, and we develop and evaluate many different submodular functions for this purpose, including determinantal point process (DPP) based approaches, submodular mutual information functions, and facility location-based functions. On a new large-scale suite of 35 heterogeneous benchmarks spanning five different capability categories, 18 frontier LLMs, and over 61K prompts, we find that the facility location (FL) function operating exclusively on inexpensive semantic prompt embeddings preserves LLM scores better than twelve separate score-based and diversity-based baselines, across a range of coreset budgets. Moreover, we show our proposed objective is not limited to the evaluation-unsupervised regime: in the setting where only a handful of whole benchmarks must be selected and a large amount of model scores are available, the same objective matches or outperforms state-of-the-art baselines on the MMLU and MTEB leaderboards, while being substantially cheaper to compute. Together, our results suggest that submodularity, in general, is a strong and reliable tool for benchmark compression.

cs.AI↗

arXiv2024

Fast classical simulation of qubit-qudit hybrid systems

Simulating quantum circuits is a computationally intensive task that relies heavily on tensor products and matrix multiplications, which can be inefficient. Recent advancements, eliminate the need for tensor products and matrix multiplications, offering significant improvements in efficiency and parallelization. Extending these optimizations, we adopt a block-simulation methodology applicable to qubit-qudit hybrid systems. This method interprets the statevector as a collection of blocks and applies gates without computing the entire circuit unitary. Our method, a spiritual successor of the simulator QuDiet \cite{Chatterjee_2023}, utilizes this block-simulation method, thereby gaining major improvements over the simulation methods used by its predecessor. We exhibit that the proposed method is approximately 10$\times$ to 1000$\times$ faster than the state-of-the-art simulator for simulating multi-level quantum systems with various benchmark circuits.

quant-ph↗

arXiv2024

How Many Images Does It Take? Estimating Imitation Thresholds in Text-to-Image Models

Text-to-image models are trained using large datasets of image-text pairs collected from the internet. These datasets often include copyrighted and private images. Training models on such datasets enables them to generate images that might violate copyright laws and individual privacy. This phenomenon is termed imitation -- generation of images with content that has recognizable similarity to its training images. In this work we estimate the point at which a model was trained on enough instances of a concept to be able to imitate it -- the imitation threshold. We posit this question as a new problem and propose an efficient approach that estimates the imitation threshold without incurring the colossal cost of training these models from scratch. We experiment with two domains -- human faces and art styles, and evaluate four text-to-image models that were trained on three pretraining datasets. We estimate the imitation threshold of these models to be in the range of 200-700 images, depending on the domain and the model. The imitation threshold provides an empirical basis for copyright violation claims and acts as a guiding principle for text-to-image model developers that aim to comply with copyright and privacy laws. Website: https://how-many-van-goghs-does-it-take.github.io/. Code: https://github.com/vsahil/MIMETIC-2.

cs.CV↗

arXiv2024

Deep Submodular Peripteral Networks

Submodular functions, crucial for various applications, often lack practical learning methods for their acquisition. Seemingly unrelated, learning a scaling from oracles offering graded pairwise preferences (GPC) is underexplored, despite a rich history in psychometrics. In this paper, we introduce deep submodular peripteral networks (DSPNs), a novel parametric family of submodular functions, and methods for their training using a GPC-based strategy to connect and then tackle both of the above challenges. We introduce newly devised GPC-style ``peripteral'' loss which leverages numerically graded relationships between pairs of objects (sets in our case). Unlike traditional contrastive learning, or RHLF preference ranking, our method utilizes graded comparisons, extracting more nuanced information than just binary-outcome comparisons, and contrasts sets of any size (not just two). We also define a novel suite of automatic sampling strategies for training, including active-learning inspired submodular feedback. We demonstrate DSPNs' efficacy in learning submodularity from a costly target submodular function and demonstrate its superiority both for experimental design and online streaming applications.

cs.LG↗

arXiv2023

FragQC: An Efficient Quantum Error Reduction Technique using Quantum Circuit Fragmentation

Quantum computers must meet extremely stringent qualitative and quantitative requirements on their qubits in order to solve real-life problems. Quantum circuit fragmentation techniques divide a large quantum circuit into a number of sub-circuits that can be executed on the smaller noisy quantum hardware available. However, the process of quantum circuit fragmentation involves finding an ideal cut that has exponential time complexity, and also classical post-processing required to reconstruct the output. In this paper, we represent a quantum circuit using a weighted graph and propose a novel classical graph partitioning algorithm for selecting an efficient fragmentation that reduces the entanglement between the sub-circuits along with balancing the estimated error in each sub-circuit. We also demonstrate a comparative study over different classical and quantum approaches of graph partitioning for finding such a cut. We present {\it FragQC}, a software tool that cuts a quantum circuit into sub-circuits when its error probability exceeds a certain threshold. With this proposed approach, we achieve an increase of fidelity by 14.83\% compared to direct execution without cutting the circuit, and 8.45\% over the state-of-the-art ILP-based method, for the benchmark circuits.

quant-ph↗

arXiv2023

Suppression of the inclination instability in the trans-Neptunian Solar system

The trans-Neptunian scattered disk exhibits unexpected dynamical structure, ranging from an extended dispersion of perihelion distance to a clustered distribution in orbital angles. Self-gravitational modulation of the scattered disk has been suggested in the literature as an alternative mechanism to Planet 9 for sculpting the orbital architecture of the trans-Neptunian region. The numerics of this hypothesis have hitherto been limited to $N < O(10^3)$ super-particle simulations that omit direct gravitational perturbations from the giant planets and instead model them as an orbit-averaged (quadrupolar) potential, through an enhanced $J_2$ moment of the central body. For sufficiently massive disks, such simulations reveal the onset of collective dynamical behaviour $\unicode{x2014}$ termed the $\unicode{x2018}$inclination instability$\unicode{x2019}$ $\unicode{x2014}$ wherein orbital circularisation occurs at the expense of coherent excitation of the inclination. Here, we report $N = O(10^4)$ GPU-accelerated simulations of a self-gravitating scattered disk (across a range of disk masses spanning 5 to 40 Earth masses) that self-consistently account for intra-particle interactions as well as Neptune's perturbations. Our numerical experiments show that even under the most favourable conditions, the inclination instability never ensues. Instead, due to scattering, the disk depletes. While our calculations show that a transient lopsided structure can emerge within the first few hundreds of Myr, the terminal outcomes of these calculations systematically reveal a scattered disk that is free of any orbital clustering. We conclude thus that the inclination instability mechanism is an inadequate explanation of the observed architecture of the solar system.

astro-ph.EP↗

arXiv2023

Accelerating Batch Active Learning Using Continual Learning Techniques

A major problem with Active Learning (AL) is high training costs since models are typically retrained from scratch after every query round. We start by demonstrating that standard AL on neural networks with warm starting fails, both to accelerate training and to avoid catastrophic forgetting when using fine-tuning over AL query rounds. We then develop a new class of techniques, circumventing this problem, by biasing further training towards previously labeled sets. We accomplish this by employing existing, and developing novel, replay-based Continual Learning (CL) algorithms that are effective at quickly learning the new without forgetting the old, especially when data comes from an evolving distribution. We call this paradigm Continual Active Learning (CAL). We show CAL achieves significant speedups using a plethora of replay schemes that use model distillation and that select diverse, uncertain points from the history. We conduct experiments across many data domains, including natural language, vision, medical imaging, and computational biology, each with different neural architectures and dataset sizes. CAL consistently provides a 3x reduction in training time, while retaining performance.

cs.LG↗

arXiv2022

QuDiet: A Classical Simulation Platform for Qubit-Qudit Hybrid Quantum Systems

In the recent years, numerous research advancements have extended the limit of classical simulation of quantum algorithms. Although, most of the state-of-the-art classical simulators are only limited to binary quantum systems, which restrict the classical simulation of higher-dimensional quantum computing systems. Through recent developments in higher-dimensional quantum computing systems, it is realized that implementing qudits improves the overall performance of a quantum algorithm by increasing memory space and reducing the asymptotic complexity of a quantum circuit. Hence, in this article, we introduce \textbf{QuDiet}, a state-of-the-art user-friendly python-based higher-dimensional quantum computing simulator. \textbf{QuDiet} offers multi-valued logic operations by utilizing generalized quantum gates with an abstraction so that any naive user can simulate qudit systems with ease as compared to the existing ones. We simulate various benchmark quantum circuits in \textbf{QuDiet} and show the considerable speedup in simulation time as compared to the other simulators without loss in precision. Finally, \textbf{QuDiet} provides a full qubit-qudit hybrid quantum simulator package with quantum circuit templates of well-known quantum algorithms for fast prototyping and simulation. The complete code and packages of \textbf{QuDiet} is available at https://github.com/LegacYFTw/QuDiet so that other platforms can incorporate it as a classical simulation option for qubit-qudit hybrid systems to their platforms.

quant-ph↗

arXiv2022

High Resolution Point Clouds from mmWave Radar

This paper explores a machine learning approach for generating high resolution point clouds from a single-chip mmWave radar. Unlike lidar and vision-based systems, mmWave radar can operate in harsh environments and see through occlusions like smoke, fog, and dust. Unfortunately, current mmWave processing techniques offer poor spatial resolution compared to lidar point clouds. This paper presents RadarHD, an end-to-end neural network that constructs lidar-like point clouds from low resolution radar input. Enhancing radar images is challenging due to the presence of specular and spurious reflections. Radar data also doesn't map well to traditional image processing techniques due to the signal's sinc-like spreading pattern. We overcome these challenges by training RadarHD on a large volume of raw I/Q radar data paired with lidar point clouds across diverse indoor settings. Our experiments show the ability to generate rich point clouds even in scenes unobserved during training and in the presence of heavy smoke occlusion. Further, RadarHD's point clouds are high-quality enough to work with existing lidar odometry and mapping workflows.

cs.RO↗

arXiv2021

Qurzon: A Prototype for a Divide and Conquer Based Quantum Compiler

When working with algorithms on quantum devices, quantum memory becomes a crucial bottleneck due to low qubit count in NISQ-era devices. In this context, the concept of `divide and compute', wherein a quantum circuit is broken into several subcircuits and executed separately, while stitching the results of the circuits via classical post-processing, becomes a viable option, especially in NISQ-era devices. This paper introduces \textbf{Qurzon}, a proposed novel quantum compiler that incorporates the marriage of techniques of divide and compute with the state-of-the-art algorithms of optimal qubit placement for executing on real quantum devices. A scheduling algorithm is also introduced within the compiler that can explore the power of distributed quantum computing while paving the way for quantum parallelism for large algorithms. Several benchmark circuits have been executed using the compiler, thereby demonstrating the power of the divide and compute when working with real NISQ-era quantum devices.

quant-ph↗

arXiv2021

A near-term quantum simulation of the transverse field Ising model hints at Glassy Dynamics

We demonstrate quantum circuit simulations of the transverse field Ising model with longitudinal fields, displaying salient features of glassy dynamics. The energy landscape and spin configurations of toy models are considered, using the Variational Quantum Eigensolver (VQE) to obtain the ground-state energies and corresponding eigenstates for a $6 \times 6$ Ising lattice using 36 qubits and a 1-dimensional Ising chain of length 25 using 25 qubits. The former showed disordered spin configurations for a specific mixture of values of the two fields. These insights mirror catalytic processes, where disorder within a catalyst can lead to inefficient reaction mechanisms. Results obtained from our proof-of-principle implementation make the case for kick-starting more concentrated efforts in harnessing existing quantum computational tools for computationally probing complex dynamical behaviour arising in quantum matter. Our aim is to leverage tools from quantum information processing to bring about a more nuanced understanding of the dynamics and structure of glassy systems, ultimately informing the development of novel materials and technology.

quant-ph↗

arXiv2020

Physics-inspired deep learning to characterize the signal manifold of quasi-circular, spinning, non-precessing binary black hole mergers

The spin distribution of binary black hole mergers contains key information concerning the formation channels of these objects, and the astrophysical environments where they form, evolve and coalesce. To quantify the suitability of deep learning to characterize the signal manifold of quasi-circular, spinning, non-precessing binary black hole mergers, we introduce a modified version of WaveNet trained with a novel optimization scheme that incorporates general relativistic constraints of the spin properties of astrophysical black holes. The neural network model is trained, validated and tested with 1.5 million $\ell=|m|=2$ waveforms generated within the regime of validity of NRHybSur3dq8, i.e., mass-ratios $q\leq8$ and individual black hole spins $ | s^z_{\{1,\,2\}} | \leq 0.8$. Using this neural network model, we quantify how accurately we can infer the astrophysical parameters of black hole mergers in the absence of noise. We do this by computing the overlap between waveforms in the testing data set and the corresponding signals whose mass-ratio and individual spins are predicted by our neural network. We find that the convergence of high performance computing and physics-inspired optimization algorithms enable an accurate reconstruction of the mass-ratio and individual spins of binary black hole mergers across the parameter space under consideration. This is a significant step towards an informed utilization of physics-inspired deep learning models to reconstruct the spin distribution of binary black hole mergers in realistic detection scenarios.

gr-qc↗

arXiv2020

EinsteinPy: A Community Python Package for General Relativity

This paper presents EinsteinPy (version 0.3), a community-developed Python package for gravitational and relativistic astrophysics. Python is a free, easy to use a high-level programming language which has seen a huge expansion in the number of its users and developers in recent years. Specifically, a lot of recent studies show that the use of Python in Astrophysics and general physics has increased exponentially. We aim to provide a very high level of abstraction, an easy to use interface and pleasing user experience. EinsteinPy is developed keeping in mind the state of a theoretical gravitational physicist with little or no background in computer programming and trying to work in the field of numerical relativity or trying to use simulations in their research. Currently, EinsteinPy supports simulation of time-like and null geodesics and calculates trajectories in different background geometries some of which are Schwarzschild, Kerr, and KerrNewmann along with coordinate inter-conversion pipeline. It has a partially developed pipeline for plotting and visualization with dependencies on libraries like Plotly, matplotlib, etc. One of the unique features of EinsteinPy is a sufficiently developed symbolic tensor manipulation utilities which are a great tool in itself for teaching yourself tensor algebra which for many beginner students can be overwhelmingly tricky. EinsteinPy also provides few utility functions for hypersurface embedding of Schwarzschild spacetime which further will be extended to model gravitational lensing simulation.

gr-qc↗