arXiv subjects

Dhruv Gupta

Publications and source records attributed to Dhruv Gupta.

14 recordsLinked to original sources

arXiv2026

Subgroup performance analysis of adaptation strategies for chest X-ray foundation models

Foundation models are increasingly adapted for downstream medical imaging tasks, yet the influence of the chosen adaptation strategy on subgroup fairness remains poorly understood. We investigate how three parameter-efficient adaptation techniques, including linear heads on the raw CLS token, an MLP, and an attention-pooling module over multi-layer patch features, affect both pathology classification performance and subgroup disparities when applied to the frozen Rad-DINO chest X-ray encoder. Using MIMIC-CXR, we evaluate eight pathologies across race, sex, and imaging-view subgroups on a prevalence-preserving, demographically balanced test set, and additionally probe how strongly each adapter encodes protected attributes. We find that attention pooling achieves the strongest overall discriminative performance and encodes attributes, particularly race, most strongly, but that improved overall performance does not consistently reduce subgroup disparities. Notably, stronger attribute encoding did not correspond to larger disparities: early network layers encoded race most weakly yet produced the largest subgroup performance gaps. Exploring different attention-pooling layer combinations further revealed no consistent relationship between the layers pooled, attribute encoding strength, and subgroup fairness. Our results indicate that richer, more expressive representations can improve accuracy while leaving fairness implications task-dependent and unpredictable, which must be assessed directly and per-task rather than inferred from encoding strength or overall performance alone.

cs.CV↗

arXiv2026

Signed Compression Progress on a Sealed Audit is Goodhart-Resistant

Compression progress is a long-standing proposal for intrinsic motivation: reward an agent when its world model becomes better at predicting or compressing experience. The folk claim is that this reward is "credible" because it is paid only for learning. We make this precise and prove it. If intrinsic reward is the signed decrease of a fixed sealed-audit loss, r_t = E(theta_{t-1}) - E(theta_t), then cumulative reward telescopes exactly to endpoint audit improvement, so no policy can push reward up indefinitely while true audit performance stagnates or degrades. For finite audit panels the same result holds with a sharp false-positive budget: cumulative empirical reward is at most true audit improvement plus 2 Delta_n(F, delta), the uniform audit deviation of the model class. This is horizon-free: adaptivity over time costs nothing once the sealed panel uniformly controls the class. The theorem also identifies the failure modes: the guarantee disappears if progress is clipped, scored on the agent's own stream, exposed to a high-capacity model on a reusable panel, or applied to a neural class that makes Delta_n vacuous. We give a Lean 4 mechanization of the structural core (telescoping, the finite-audit bound, finite Gibbs, and the entropy floor) and an experiment suite on ARC-TGI grid-transformation generators with adaptive holdout attacks. Experiments confirm the theory: finite-audit deviation scales as n^{-0.527}; signed progress resists clip-farming, stream leakage, and noisy-TV curiosity; naive reusable audits are exploitable by black-box scalar feedback, while standard release defenses keep the attack below the 2 Delta_n threshold. Signed compression progress on a sealed audit is an accounting signal of genuine improvement.

cs.LG↗

arXiv2026

Null Measurability at the Symmetrization Interface in VC Learning

Recent work revisiting measurability in the fundamental theorem of statistical learning imposes Borel measurability of ghost-gap suprema. We show that, at the one-sided ghost-gap interface actually used by the standard symmetrization proof, this requirement is stronger than necessary. For any Borel-parameterized concept class on a Polish domain, the bad event "there exists a hypothesis whose ghost empirical error exceeds its training empirical error by at least {\epsilon}/2" is analytic. By Choquet capacitability, it is therefore measurable in the completion of every finite Borel measure. We then construct a concept class whose bad event is null-measurable but not Borel, giving a strict separation from the Borel supremum condition. Finally, we prove closure under patching, fixed and countable interpolation, and fiber-product amalgamation, showing that the weaker regularity level is stable under natural concept-class constructors. In the realizable setting, where targets belong to the class and are measurable, these results weaken the measurability hypothesis needed by the symmetrization route from finite VC dimension to PAC learnability. The main results and the descriptive-set-theoretic infrastructure used by them are formalized in Lean 4.

cs.LG↗

arXiv2026

The solvability of the inverse volcano problem over non-prime finite fields

For a finite field $\mathbf{F}_{p^k}$ and a prime $\ell \neq p$, consider the graph $G$ of $\ell$-isogenies between ordinary elliptic curves over $\mathbf{F}_{p^k}$. Kohel proved that the connected components of $G$ have a remarkable structure, now called an $\ell$-volcano graph. Bambury, Campagna, and Pazuki investigated the inverse volcano problem: given a volcano graph $V$, can one find it as a connected component of $G$ over $\mathbf{F}_{p^k}$? They gave a complete positive answer over $\mathbf{F}_p$, and described a specific counterexample over $\mathbf{F}_{p^2}$. In this paper, we generalise the results of Bambury-Campagna-Pazuki by providing a precise framework for the inverse volcano problem over $\mathbf{F}_{p^k}$. The solvability of the problem for an $\ell$-volcano graph $V$ of depth $d$ is typically determined by the relation between $d$ and the $\ell$-valuation $r$ of $k$. When $r$ is small in comparison to $d$, we prove that there are infinitely many primes $p$ solving the inverse problem for $V$. The situation where $r$ is large in comparison to $d$ is more delicate: in many cases we prove that the inverse problem for $V$ is unsolvable; in a few other cases the problem appears to be solvable, but our proof of this is conditional on a variant of the Cohen-Lenstra heuristics for class groups of imaginary quadratic fields. We provide some computational evidence in support of these modified heuristics.

math.NT↗

arXiv2025

Do You Really Need a GPU to Guard Your LLM? CPU-Class Classifiers and Multi-Stage Pipelines for Safety Enforcement at Scale

Safety classifiers that screen LLM inputs for jailbreak attempts have become standard deployment components, yet almost all production systems rely on GPU-based models: fine-tuned transformers and LLM-as-a-judge pipelines. These approaches impose significant per-query latency and infrastructure cost. Very little research has asked whether CPU-based classifiers, such as support vector machines and gradient-boosted trees trained on TF-IDF features, can match their accuracy across the conditions that production deployments encounter. We evaluate five CPU classifier families, Mamba-130M as an SSM-based GPU classifier, and transformer-based GPU models (DeBERTa-v3 and Gemma-2B with LoRA) across nine jailbreak sources and three regimes: in-distribution (D1), out-of-distribution (D2), and adversarially obfuscated (D3). On D1, the best CPU classifier matches the best transformer GPU model at roughly one-fifth the deployment cost. On D2, CPU classifiers fail via confident miscalibration, producing high-confidence false negatives that bypass escalation entirely. On D3, CPU classifiers outperform transformer GPU models by more than 26 percentage points in F1. Based on these complementary failure modes, we design GuardChain, a three-stage safety pipeline (Regex -> CPU -> GPU) that routes each prompt to the cheapest stage capable of a confident decision. The CPU stage alone resolves 80\% of in-distribution prompts at near-peak accuracy, and the GPU stage recovers the out-of-distribution failures. For practitioners deploying LLM safety at scale, this work provides evidence that GPU-class infrastructure is unnecessary for the majority of traffic.

cs.CR↗

arXiv2025

eDCF: Estimating Intrinsic Dimension using Local Connectivity

Modern datasets often contain high-dimensional features exhibiting complex dependencies. To effectively analyze such data, dimensionality reduction methods rely on estimating the dataset's intrinsic dimension (id) as a measure of its underlying complexity. However, estimating id is challenging due to its dependence on scale: at very fine scales, noise inflates id estimates, while at coarser scales, estimates stabilize to lower, scale-invariant values. This paper introduces a novel, scalable, and parallelizable method called eDCF, which is based on Connectivity Factor (CF), a local connectivity-based metric, to robustly estimate intrinsic dimension across varying scales. Our method consistently matches leading estimators, achieving comparable values of mean absolute error (MAE) on synthetic benchmarks with noisy samples. Moreover, our approach also attains higher exact intrinsic dimension match rates, reaching up to 25.0% compared to 16.7% for MLE and 12.5% for TWO-NN, particularly excelling under medium to high noise levels and large datasets. Further, we showcase our method's ability to accurately detect fractal geometries in decision boundaries, confirming its utility for analyzing realistic, structured data.

cs.LG↗

arXiv2025

SACL: Understanding and Combating Textual Bias in Code Retrieval with Semantic-Augmented Reranking and Localization

Retrieval-Augmented Code Generation (RACG) is a critical technique for enhancing code generation by retrieving relevant information. In this work, we conduct an in-depth analysis of code retrieval by systematically masking specific features while preserving code functionality. Our discoveries include: (1) although trained on code, current retrievers heavily rely on surface-level textual features (e.g., docstrings, identifier names), and (2) they exhibit a strong bias towards well-documented code, even if the documentation is irrelevant. Based on our discoveries, we propose SACL, a framework that enriches textual information and reduces bias by augmenting code or structural knowledge with semantic information. Extensive experiments show that SACL substantially improves code retrieval (e.g., by 12.8% / 9.4% / 7.0% Recall@1 on HumanEval / MBPP / SWE-Bench-Lite), which also leads to better code generation performance (e.g., by 4.88% Pass@1 on HumanEval).

cs.CL↗

arXiv2025

Hybrid Deep Learning Framework for Classification of Kidney CT Images: Diagnosis of Stones, Cysts, and Tumors

Medical image classification is a vital research area that utilizes advanced computational techniques to improve disease diagnosis and treatment planning. Deep learning models, especially Convolutional Neural Networks (CNNs), have transformed this field by providing automated and precise analysis of complex medical images. This study introduces a hybrid deep learning model that integrates a pre-trained ResNet101 with a custom CNN to classify kidney CT images into four categories: normal, stone, cyst, and tumor. The proposed model leverages feature fusion to enhance classification accuracy, achieving 99.73% training accuracy and 100% testing accuracy. Using a dataset of 12,446 CT images and advanced feature mapping techniques, the hybrid CNN model outperforms standalone ResNet101. This architecture delivers a robust and efficient solution for automated kidney disease diagnosis, providing improved precision, recall, and reduced testing time, making it highly suitable for clinical applications.

eess.IV↗

arXiv2024

TNNGen: Automated Design of Neuromorphic Sensory Processing Units for Time-Series Clustering

Temporal Neural Networks (TNNs), a special class of spiking neural networks, draw inspiration from the neocortex in utilizing spike-timings for information processing. Recent works proposed a microarchitecture framework and custom macro suite for designing highly energy-efficient application-specific TNNs. These recent works rely on manual hardware design, a labor-intensive and time-consuming process. Further, there is no open-source functional simulation framework for TNNs. This paper introduces TNNGen, a pioneering effort towards the automated design of TNNs from PyTorch software models to post-layout netlists. TNNGen comprises a novel PyTorch functional simulator (for TNN modeling and application exploration) coupled with a Python-based hardware generator (for PyTorch-to-RTL and RTL-to-Layout conversions). Seven representative TNN designs for time-series signal clustering across diverse sensory modalities are simulated and their post-layout hardware complexity and design runtimes are assessed to demonstrate the effectiveness of TNNGen. We also highlight TNNGen's ability to accurately forecast silicon metrics without running hardware process flow.

cs.AR↗

arXiv2024

Movement of insurgent gangs: A Bayesian kernel density model for incomplete temporal data

We develop a Bayesian modeling framework to address a pressing real-life problem faced by the police in tackling insurgent gangs. Unlike criminals associated with common crimes such as robbery, theft or street crime, insurgent gangs are trained in sophisticated arms and strategise against the government to weaken its resolve. They are constantly on the move, operating over large areas causing damage to national properties and terrorizing ordinary citizens. Different from the more commonly addressed problem of modeling crime-events, our context requires that an approach be formulated to model the movement of insurgent gangs, which is more valuable to the police forces in preempting their activities and nabbing them. This paper evolved as a collaborative work with the Indian police to help augment their tactics with a systematic method, by integrating past data on observed gang-locations with the expert knowledge of the police officers. A methodological challenge in modeling the movement of insurgent gangs is that the data on their locations is incomplete, since they are observable only at some irregularly separated time-points. Based on a weighted kernel density formulation for temporal data, we analytically derive the closed form of the likelihood, conditional on incomplete past observed data. Building on the current tactics used by the police, we device an approach for constructing an expert-prior on gang-locations, along with a sequential Bayesian procedure for estimation and prediction. We also propose a new metric for predictive assessment that complements another known metric used in similar problems.

stat.AP↗

arXiv2023

Evolving Domain Adaptation of Pretrained Language Models for Text Classification

Adapting pre-trained language models (PLMs) for time-series text classification amidst evolving domain shifts (EDS) is critical for maintaining accuracy in applications like stance detection. This study benchmarks the effectiveness of evolving domain adaptation (EDA) strategies, notably self-training, domain-adversarial training, and domain-adaptive pretraining, with a focus on an incremental self-training method. Our analysis across various datasets reveals that this incremental method excels at adapting PLMs to EDS, outperforming traditional domain adaptation techniques. These findings highlight the importance of continually updating PLMs to ensure their effectiveness in real-world applications, paving the way for future research into PLM robustness against the natural temporal evolution of language.

cs.CL↗

arXiv2020

Generating Fact Checking Summaries for Web Claims

We present SUMO, a neural attention-based approach that learns to establish the correctness of textual claims based on evidence in the form of text documents (e.g., news articles or Web documents). SUMO further generates an extractive summary by presenting a diversified set of sentences from the documents that explain its decision on the correctness of the textual claim. Prior approaches to address the problem of fact checking and evidence extraction have relied on simple concatenation of claim and document word embeddings as an input to claim driven attention weight computation. This is done so as to extract salient words and sentences from the documents that help establish the correctness of the claim. However, this design of claim-driven attention does not capture the contextual information in documents properly. We improve on the prior art by using improved claim and title guided hierarchical attention to model effective contextual cues. We show the efficacy of our approach on datasets concerning political, healthcare, and environmental issues.

cs.CL↗

arXiv2016

Event Search and Analytics: Detecting Events in Semantically Annotated Corpora for Search and Analytics

In this article, I present the questions that I seek to answer in my PhD research. I posit to analyze natural language text with the help of semantic annotations and mine important events for navigating large text corpora. Semantic annotations such as named entities, geographic locations, and temporal expressions can help us mine events from the given corpora. These events thus provide us with useful means to discover the locked knowledge in them. I pose three problems that can help unlock this knowledge vault in semantically annotated text corpora: i. identifying important events; ii. semantic search; and iii. event analytics.

cs.IR↗

arXiv2015

Driver Friendly Headlight Controller for Driving in Developing Countries

In developing countries, night driving is extremely unsafe mainly due to; 1). Poorly followed traffic rules and 2) bad road conditions. The number of accidents is increasing at a frightening pace, necessitating the development of a low cost automatic headlight control system to ensure safety. In most accident cases, fatal collisions take place due to glare generated by excessive headlight intensity (high beam) of the oncoming vehicle. In this paper, a user friendly controller for headlight intensity control for driving on highway at night has been proposed. Aim is to design simple and affordable system that can alleviate effect of blind spot due to high glare on the vehicle windscreen. Controller is based on Fuzzy inference system (FIS) and used incoming light intensity as criteria. Also, relative distance and speed is derived using incoming intensity of the oncoming vehicles. The system is designed considering human tolerance levels of light intensity as the boundary values. The system controls headlight beam voltage such that; a) intensity is maintained in human visual comfort zone and b) blind spot is avoided. The super user feature incorporated in the intensity controller offers personalized preferences to the driver. Due to the use of the single parameter (intensity) for control, system can be implemented using simplified hardware.

eess.SY↗