arXiv subjects

Nikhil Garg

Publications and source records attributed to Nikhil Garg.

At least 19 recordsLinked to original sources

arXiv2026

Hitting a Moving Target: Test-Time Adaptation for AI Text Detection under Continual Distribution Shift

Deployed approaches for AI text detection often rely on training-time access to labeled datasets of both human-written and AI-generated text. This approach is vulnerable to three types of distribution shifts that occur continually post-deployment, and for which labeled data is often unavailable: adversarial humanization, new LLMs being released, and temporal drift in human writing. Simultaneously, existing approaches do not leverage a key signal of LLM usage: inference-time homogeneity. We propose a test-time adaptation (TTA) approach, using semi-supervised learning, that adapts to distribution shifts by leveraging homogeneity among unlabeled samples observed at inference time. Empirically, we find that state-of-the-art supervised detectors systematically fail when they encounter distribution shifts in AI-generated and human writing, both adversarial and natural, while test-time adaptation with semi-supervised learning is largely robust; e.g., the commercial model Pangram detects just 24.1% of our adversarial AI-generated text, compared to 90.5% for our test-time approach. We establish that test-time adaptation is a promising framework for AI text detection in the wild. We publicly release our code (which includes code for model training, evaluation, and plots) at https://github.com/kkr36/llm_detection.

cs.CL↗

arXiv2026

EconCSLib: AI-Assisted Lean Formalization for Economics & Computation research

This paper presents EconCSLib, a Lean 4 library and workflow for formalizing research papers in applied modeling fields such as Economics and Computation, with language-model assistance. The goal of EconCSLib is to enable researchers to formalize their papers in Lean without knowing Lean themselves. The central design principle is a human-AI-Lean formalization workflow: an LLM writes Lean code, Lean checks formal statements and proofs, and both humans and LLM-as-judge processes can verify that the paper's statements were translated into Lean correctly. We develop agent skills, human-facing reporting, a review dashboard, and auditing procedures to support this workflow. The current public repository contains 20 formalized papers and 4 partially formalized papers, along with shared libraries for probability (including stochastic processes), auctions, matching markets, social choice, and graph tools, totaling 986,391 lines of Lean code. To our knowledge, we are also among the first applied math researchers to systematically pursue Lean formalization of one's own publications in the process of building such a community library. We welcome users and contributors to the project. The library and workflow are available at https://github.com/nikhgarg/EconCSLib, with corresponding project webpage at https://gargnikhil.com/EconCSLib/.

cs.GT↗

arXiv2026

The Subjectivity of Monoculture

Machine learning models -- including large language models (LLMs) -- are often said to exhibit monoculture, where outputs agree strikingly often. But what does it actually mean for models to agree too much? We argue that this question is inherently subjective, relying on two key decisions. First, the analyst must specify a baseline null model for what "independence" should look like. This choice is inherently subjective, and as we show, different null models result in dramatically different inferences about excess agreement. Second, we show that inferences depend on the population of models and items under consideration. Models that seem highly correlated in one context may appear independent when evaluated on a different set of questions, or against a different set of peers. Experiments on two large-scale benchmarks validate our theoretical findings. For example, we find drastically different inferences when using a null model with item difficulty compared to previous works that do not. Together, our results reframe monoculture evaluation not as an absolute property of model behavior, but as a context-dependent inference problem.

cs.CY↗

arXiv2026

Simpler Than You Think: The Practical Dynamics of Ranked Choice Voting

Ranked Choice Voting (RCV) adoption is expanding across U.S. elections, but faces persistent criticism for complexity, strategic manipulation, and ballot exhaustion. We empirically test these concerns on real election data, across three diverse contexts: New York City's 2021 Democratic primaries (54 races), Alaska's 2024 primary-infused statewide elections (52 races), and Portland's 2024 multi-winner City Council elections (4 races). Our algorithmic approach circumvents computational complexity barriers by reducing election instance sizes (via candidate elimination). Our findings reveal that despite its intricate multi-round process and theoretical vulnerabilities, RCV consistently exhibits simple and transparent dynamics in practice, closely mirroring the interpretability of plurality elections. Following RCV adoption, competitiveness increased substantially compared to prior plurality elections, with average margins of victory declining by 9.2 percentage points in NYC and 11.4 points in Alaska. Empirically, complex ballot-addition strategies are not more efficient than simple ones, and ballot exhaustion has minimal impact, altering outcomes in only 3 of 110 elections. These findings demonstrate that RCV delivers measurable democratic benefits while proving robust to ballot-addition manipulation, resilient to ballot exhaustion effects, and maintaining transparent competitive dynamics in practice. The computational framework offers election administrators and researchers tools for immediate election-night analysis and facilitating clearer discourse around election dynamics.

cs.CY↗

arXiv2026

How Many Features Can a Language Model Store Under the Linear Representation Hypothesis?

We introduce a mathematical framework for the linear representation hypothesis (LRH), which asserts that intermediate layers of language models store features linearly. We separate the hypothesis into two claims: linear representation (features are linearly embedded in neuron activations) and linear accessibility (features can be linearly decoded). We then ask: How many neurons $d$ suffice to both linearly represent and linearly access $m$ features? Classical results in compressed sensing imply that for $k$-sparse inputs, $d = O(k\log (m/k))$ suffices if we allow non-linear decoding algorithms (Candes and Tao, 2006; Candes et al., 2006; Donoho, 2006). However, the additional requirement of linear decoding takes the problem out of the classical compressed sensing, into linear compressed sensing. Our main theoretical result establishes nearly-matching upper and lower bounds for linear compressed sensing. We prove that $d = \Omega_\epsilon(\frac{k^2}{\log k}\log (m/k))$ is required while $d = O_\epsilon(k^2\log m)$ suffices. The lower bound establishes a quantitative gap between classical and linear compressed setting, illustrating how linear accessibility is a meaningfully stronger hypothesis than linear representation alone. The upper bound confirms that neurons can store an exponential number of features under the LRH, giving theoretical evidence for the "superposition hypothesis" (Elhage et al., 2022). The upper bound proof uses standard random constructions of matrices with approximately orthogonal columns. The lower bound proof uses rank bounds for near-identity matrices (Alon, 2003) together with Tur\'an's theorem (bounding the number of edges in clique-free graphs). We also show how our results do and do not constrain the geometry of feature representations and extend our results to allow decoders with an activation function and bias.

cs.LG↗

arXiv2026

Capacity Constraints Make Admissions Processes Less Predictable

Machine learning models are often used to make predictions about admissions process outcomes, such as for colleges or jobs. However, such decision processes differ substantially from the conventional machine learning paradigm. Because admissions decisions are capacity-constrained, whether a student is admitted depends on the other applicants who apply. We show how this dependence affects predictive performance even in otherwise ideal settings. Theoretically, we introduce two concepts that characterize the relationship between admission function properties, machine learning representation, and generalization to applicant pool distribution shifts: instability, which measures how many existing decisions can change when a single new applicant is introduced; and variability, which measures the number of unique students whose decisions can change. Empirically, we illustrate our theory on individual-level admissions data from the New York City high school matching system, showing that machine learning performance degrades as the applicant pool increasingly differs from the training data. Furthermore, there are larger performance drops for schools using decision rules that are more unstable and variable. Our work raises questions about the reliability of predicting individual admissions probabilities.

cs.CY↗

arXiv2026

Paper Skygest: Personalized Academic Recommendations on Bluesky

We build, deploy, and evaluate Paper Skygest, a custom personalized social feed for scientific content posted by a user's network on Bluesky and the AT Protocol. We leverage a new capability on emerging decentralized social media platforms: the ability for anyone to build and deploy feeds for other users, to use just as they would a native platform-built feed. To our knowledge, Paper Skygest is the first and largest such continuously deployed personalized social media feed by academics, with over 50,000 weekly uses by over 1,000 daily active users, all organically acquired. First, we quantitatively and qualitatively evaluate Paper Skygest usage, showing that it has sustained usage and satisfies users; we further show adoption of Paper Skygest increases a user's interactions with posts about research, and how interaction rates change as a function of post order. Second, we share our full code and describe our system architecture, to support other academics in building and deploying such feeds sustainably. Third, we overview the potential of custom feeds such as Paper Skygest for studying algorithm designs, building for user agency, and running recommender system experiments with organic users without partnering with a centralized platform.

cs.SI↗

arXiv2026

Analog Weight Update Rule in Ferroelectric Hafnia, using pico-Joule Programming Pulses

In an effort to compete with the brain's efficiency at processing information, neuromorphic hardware combines artificial synapses and neurons using mixed-signal circuits and emerging memories. In ferroelectric resistive weights, the strength of the synaptic connection between two neurons is stored in the device conductance. During learning, programming pulses are applied to the synaptic weight, which reconfigures the ferroelectric domains and adjusts the conductance. One strategy to lower the energy cost during the training phase is to lower the duration of the programming pulses. However, the latter cannot be shorter than the self-loading time of the resistive weights, limited by intrinsic parasitics in the circuits. In this work, ferroelectric resistive weights are fabricated using a process compatible with CMOS Back-End-Of-Line integration, based on hafnia/zirconia nanolaminates. By laterally scaling the device area under 100 $\mu$m$^2$, the self-loading time becomes sufficiently short to enable 20 ns programming, which corresponds to a maximum of 3 picoJoules per pulse. Further, in this work, the weight update rule with 20 ns pulses is experimentally measured not only for different amplitudes but also for different initial conductance states. We find that the final weight is determined by the pulse amplitude, independent of the initial weight value.

cs.ET↗

arXiv2025

Personalized Spiking Neural Networks with Ferroelectric Synapses for EEG Signal Processing

Electroencephalography (EEG)-based brain-computer interfaces (BCIs) are strongly affected by non-stationary neural signals that vary across sessions and individuals, limiting the generalization of subject-agnostic models and motivating adaptive and personalized learning on resource-constrained platforms. Programmable memristive hardware offers a promising substrate for such post-deployment adaptation; however, practical realization is challenged by limited weight resolution, device variability, nonlinear programming dynamics, and finite device endurance. In this work, we show that spiking neural networks (SNNs) can be deployed on ferroelectric memristive synaptic devices for adaptive EEG-based motor imagery decoding under realistic device constraints, achieving classification performance comparable to software-based SNNs. We fabricate, characterize, and model the weight update in ferroelectric synapses. We then evaluate the deployment of convolutional-recurrent SNN architecture using two strategies. First, we adapt to SNNs a mixed precision strategy in which gradient-based updates are accumulated digitally and converted into discrete programming events only when a threshold is exceeded. Additionally, the weight update is device-aware and accounts for the nonlinear, state-dependent programming dynamics. During learning and adaptation, this scheme mitigates possible endurance and energy constraints. Second, we evaluate the transfer of software-trained weights followed by low-overhead on-device re-tuning. We show that, subject-specific transfer learning achieved by retraining only the final network layers improves classification accuracy. These results demonstrate that programmable ferroelectric hardware can support robust, low-overhead adaptation in spiking neural networks, opening a practical path toward personalized neuromorphic processing of neural signals.

cs.NE↗

arXiv2025

Unsupervised local learning based on voltage-dependent synaptic plasticity for resistive and ferroelectric synapses

The deployment of AI on edge computing devices faces significant challenges related to energy consumption and functionality. These devices could greatly benefit from brain-inspired learning mechanisms, allowing for real-time adaptation while using low-power. In-memory computing with nanoscale resistive memories may play a crucial role in enabling the execution of AI workloads on these edge devices. In this study, we introduce voltage-dependent synaptic plasticity (VDSP) as an efficient approach for unsupervised and local learning in memristive synapses based on Hebbian principles. This method enables online learning without requiring complex pulse-shaping circuits typically necessary for spike-timing-dependent plasticity (STDP). We show how VDSP can be advantageously adapted to three types of memristive devices (TiO$_2$, HfO$_2$-based metal-oxide filamentary synapses, and HfZrO$_4$-based ferroelectric tunnel junctions (FTJ)) with disctinctive switching characteristics. System-level simulations of spiking neural networks incorporating these devices were conducted to validate unsupervised learning on MNIST-based pattern recognition tasks, achieving state-of-the-art performance. The results demonstrated over 83% accuracy across all devices using 200 neurons. Additionally, we assessed the impact of device variability, such as switching thresholds and ratios between high and low resistance state levels, and proposed mitigation strategies to enhance robustness.

cs.NE↗

arXiv2025

Energy-convergence trade off for the training of neural networks on bio-inspired hardware

The increasing deployment of wearable sensors and implantable devices is shifting AI processing demands to the extreme edge, necessitating ultra-low power for continuous operation. Inspired by the brain, emerging memristive devices promise to accelerate neural network training by eliminating costly data transfers between compute and memory. Though, balancing performance and energy efficiency remains a challenge. We investigate ferroelectric synaptic devices based on HfO2/ZrO2 superlattices and feed their experimentally measured weight updates into hardware-aware neural network simulations. Across pulse widths from 20 ns to 0.2 ms, shorter pulses lower per-update energy but require more training epochs while still reducing total energy without sacrificing accuracy. Classification accuracy using plain stochastic gradient descent (SGD) is diminished compared to mixed-precision SGD. We analyze the causes and propose a ``symmetry point shifting'' technique, addressing asymmetric updates and restoring accuracy. These results highlight a trade-off among accuracy, convergence speed, and energy use, showing that short-pulse programming with tailored training significantly enhances on-chip learning efficiency.

cs.ET↗

arXiv2025

Heterogeneous participation and allocation skews: when is choice "worth it"?

A core ethos of the Economics and Computation (EconCS) community is that people have complex private preferences and information of which the central planner is unaware, but which an appropriately designed mechanism can uncover to improve collective decisionmaking. This ethos underlies the community's largest deployed success stories, from stable matching systems to participatory budgeting. I ask: is this choice and information aggregation ``worth it''? In particular, I discuss how such systems induce \textit{heterogeneous participation}: those already relatively advantaged are, empirically, more able to pay time costs and navigate administrative burdens imposed by the mechanisms. I draw on three case studies, including my own work -- complex democratic mechanisms, resident crowdsourcing, and school matching. I end with lessons for practice and research, challenging the community to help reduce participation heterogeneity and design and deploy mechanisms that meet a ``best of both worlds'' north star: \textit{use preferences and information from those who choose to participate, but provide a ``sufficient'' quality of service to those who do not.}

cs.CY↗

arXiv2025

Position: Use Sparse Autoencoders to Discover Unknowns

While sparse autoencoders (SAEs) have generated significant excitement, a series of negative results have added to skepticism about their usefulness. Here, we establish a conceptual distinction that reconciles competing narratives surrounding SAEs. We argue that even if SAEs may be less effective for \textit{acting on known concepts}, SAEs are especially powerful tools for \textit{discovering unknown concepts}. This distinction separates existing negative results from positive results, and suggests several classes of SAE applications. Specifically, we outline use cases for SAEs in (i) ML interpretability, explainability, fairness, auditing, and safety, and (ii) social and health sciences.

cs.LG↗

arXiv2025

Urban Incident Prediction with Graph Neural Networks: Integrating Government Ratings and Crowdsourced Reports

Graph neural networks (GNNs) are widely used in urban spatiotemporal forecasting, such as predicting infrastructure problems. In this setting, government officials wish to know in which neighborhoods incidents like potholes or rodent issues occur. The true state of incidents (e.g., street conditions) for each neighborhood is observed via government inspection ratings. However, these ratings are only conducted for a sparse set of neighborhoods and incident types. We also observe the state of incidents via crowdsourced reports, which are more densely observed but may be biased due to heterogeneous reporting behavior. First, for such settings, we propose a multiview, multioutput GNN-based model that uses both unbiased rating data and biased reporting data to predict the true latent state of incidents. Second, we investigate a case study of New York City urban incidents and collect, standardize, and make publicly available a dataset of 9,615,863 crowdsourced reports and 1,041,415 government inspection ratings over 3 years and across 139 types of incidents. Finally, we show on both real and semi-synthetic data that our model can better predict the latent state compared to models that use only reporting data or models that use only rating data, especially when rating data is sparse and reports are predictive of ratings. We also quantify demographic biases in crowdsourced reporting, e.g., higher-income neighborhoods report problems at higher rates. Our analysis showcases a widely applicable approach for latent state prediction using heterogeneous, sparse, and biased data.

cs.LG↗

arXiv2025

Correlated Errors in Large Language Models

Diversity in training data, architecture, and providers is assumed to mitigate homogeneity in LLMs. However, we lack empirical evidence on whether different LLMs differ meaningfully. We conduct a large-scale empirical evaluation on over 350 LLMs overall, using two popular leaderboards and a resume-screening task. We find substantial correlation in model errors -- on one leaderboard dataset, models agree 60% of the time when both models err. We identify factors driving model correlation, including shared architectures and providers. Crucially, however, larger and more accurate models have highly correlated errors, even with distinct architectures and providers. Finally, we show the effects of correlation in two downstream tasks: LLM-as-judge evaluation and hiring -- the latter reflecting theoretical predictions regarding algorithmic monoculture.

cs.CL↗

arXiv2025

Optimizing Library Usage and Browser Experience: Application to the New York Public Library

We tackle the challenge brought to urban library systems by the {holds system} -- which allows users to request books available at other branches to be transferred for local pickup. The holds system increases usage of the entire collection, at the expense of an in-person browser's experience at the source branch. We study the optimization of usage and browser experience, where the library has two levers: where a book should come from when a hold request is placed, and how many book copies at each branch should be available through the holds system versus reserved for browsers. We first show that the problem of maximizing usage can be viewed through the lens of revenue management, for which near-optimal fulfillment policies exist. We then develop a simulation framework that further optimizes for browser experience, through book reservations. We empirically apply our methods to data from the New York Public Library to design implementable policies. We find that though a substantial trade-off exists between these two desiderata, a balanced policy can improve browser experience over the historical policy without significantly sacrificing usage. Because browser usage is more prevalent among branches in low-income areas, this policy further increases system-wide equity: notably, for branches in the 25% lowest-income neighborhoods, it improves both usage and browser experience by about 15%.

cs.CY↗

arXiv2025

Inferring fine-grained migration patterns across the United States

Fine-grained migration data illuminate demographic, environmental, and health phenomena. However, United States migration data have serious drawbacks: public data lack spatial granularity, and higher-resolution proprietary data suffer from multiple biases. To address this, we develop a method that fuses high-resolution proprietary data with coarse Census data to create MIGRATE: annual migration matrices capturing flows between 47.4 billion US Census Block Group pairs -- approximately four thousand times the spatial resolution of current public data. Our estimates are highly correlated with external ground-truth datasets and improve accuracy relative to raw proprietary data. We use MIGRATE to analyze national and local migration patterns. Nationally, we document demographic and temporal variation in homophily, upward mobility, and moving distance -- for example, rising moves into top-income-quartile block groups and racial disparities in upward mobility. Locally, MIGRATE reveals patterns such as wildfire-driven out-migration that are invisible in coarser previous data. We release MIGRATE as a resource for migration researchers.

cs.CY↗

arXiv2025

Bayesian Modeling of Zero-Shot Classifications for Urban Flood Detection

Street scene datasets, collected from Street View or dashboard cameras, offer a promising means of detecting urban objects and incidents like street flooding. However, a major challenge in using these datasets is their lack of reliable labels: there are myriad types of incidents, many types occur rarely, and ground-truth measures of where incidents occur are lacking. Here, we propose BayFlood, a two-stage approach which circumvents this difficulty. First, we perform zero-shot classification of where incidents occur using a pretrained vision-language model (VLM). Second, we fit a spatial Bayesian model on the VLM classifications. The zero-shot approach avoids the need to annotate large training sets, and the Bayesian model provides frequent desiderata in urban settings - principled measures of uncertainty, smoothing across locations, and incorporation of external data like stormwater accumulation zones. We comprehensively validate this two-stage approach, showing that VLMs provide strong zero-shot signal for floods across multiple cities and time periods, the Bayesian model improves out-of-sample prediction relative to baseline methods, and our inferred flood risk correlates with known external predictors of risk. Having validated our approach, we show it can be used to improve urban flood detection: our analysis reveals 113,738 people who are at high risk of flooding overlooked by current methods, identifies demographic biases in existing methods, and suggests locations for new flood sensors. More broadly, our results showcase how Bayesian modeling of zero-shot LM annotations represents a promising paradigm because it avoids the need to collect large labeled datasets and leverages the power of foundation models while providing the expressiveness and uncertainty quantification of Bayesian models.

cs.LG↗