arXiv subjects

Ravi Kumar

Publications and source records attributed to Ravi Kumar.

At least 37 records · Page 2Linked to original sources

arXiv2026

Sketching Intersection Profiles: A Simple Proof and Three Applications

In this work we settle the complexity of three sketching problems. (i) We show that sketching vertex neighborhood sizes in graphs requires $Ω(n^2)$ bits, standing in sharp contrast to the $\tilde{O}(n)$ complexity of sketching edge cuts. (ii) We obtain tight lower and upper bounds of $\tildeΘ(n^2)$ for sketching coverage functions with additive and multiplicative errors. (iii) We prove an $Ω(n^2)$ lower bound for sketching Random Utility Models under the $\ell_\infty$-norm, improving upon the previous $Ω(n \log n)$ bound and matching a known upper bound to within logarithmic factors. These bounds are obtained through a connection with the problem of sketching the intersection profile of a distribution $D$ on $2^{[n]}$. Specifically, we seek a succinct data structure that, for any query set $S \subseteq [n]$, approximates the quantity $\Pr_{T \sim D}[T \cap S \neq \varnothing]$ to within a small constant additive error. One can obtain lower bounds for this latter problem directly from known results about the itemset frequency estimation problem in databases for which tight bounds are known. As an additional contribution, we also provide an alternative proof for the intersection profile sketching lower bound, in the setting in which the accuracy parameter is constant. This proof relies solely on elementary probability avoiding the heavier machinery used in previous proofs.

cs.DS↗

arXiv2026

On the LSH Distortion of Ulam and Cayley Similarities

Locality-sensitive hashing (LSH) has found widespread use as a fundamental primitive, particularly to accelerate nearest neighbor search. An LSH scheme for a similarity function $S:\mathcal{X} \times \mathcal{X} \to [0,1]$ is a distribution over hash functions on $\mathcal{X}$ with the property that the probability of collision of any two elements $x,y\in \mathcal{X}$ is exactly equal to $S(x,y)$. However, not all similarity functions admit exact LSH schemes. The notion of LSH distortion measures how multiplicatively close a similarity function is to having an LSH scheme. In this work, we study the LSH distortion of the Ulam and Cayley similarities, which are popular similarity measures on permutations of $n$ elements. We show that the Ulam similarity admits a sublinear LSH distortion of $O(n / \sqrt{\log n})$; we also prove a lower bound of $Ω(n^{0.12})$ on the best LSH distortion achievable. On the other hand, we show that the LSH distortion of the Cayley similarity is $Θ(n)$.

cs.DS↗

arXiv2026

Convex Optimization with Local Label Differential Privacy: Tight Bounds in All Privacy Regimes

We study the problem of Stochastic Convex Optimization (SCO) under the constraint of local Label Differential Privacy (L-LDP). In this setting, the features are considered public, but the corresponding labels are sensitive and must be randomized by each user locally before being sent to an untrusted analyzer. Prior work for SCO under L-LDP (Ghazi et al., 2021) established an excess population risk bound with a \emph{linear} dependence on the size of the label space, $K$: $O\left({\frac{K}{ε\sqrt{n}}}\right)$ in the high-privacy regime ($ε\leq 1$) and $O\left({\frac{K}{e^ε \sqrt{n}}}\right)$ in the medium-privacy regime ($1 \leq ε\leq \ln K$). This left open whether this linear cost is fundamental to the L-LDP model. In this note, we resolve this question. First, we present a novel and efficient non-interactive L-LDP algorithm that achieves an excess risk of $O\left({\sqrt{\frac{K}{εn}}}\right)$ in the high-privacy regime ($ε\leq 1$) and $O\left({\sqrt{\frac{K}{e^ε n}}}\right)$ in the medium-privacy regime ($1 \leq ε\leq \ln K$). This quadratically improves the dependency on the label space size from $O(K)$ to $O(\sqrt{K})$. Second, we prove a matching information-theoretic lower bound across all privacy regimes for any sufficiently large $n$.

cs.DS↗

arXiv2026

Stabilisation of NV centres in diamond nanopillars at low temperature

Degradation of near surface nitrogen vacancy (NV) centers in diamond under optical illumination has restricted their deployment in applications such as scanning NV magnetomety, particularly under harsh environment such as low temperatures and vacuum. Previously, alumina passivation of planar diamond samples has been shown to reduce the degradation of near surface ensemble NV centers in vacuum. Here, we expand this study to incorporate photonic nanostructures by analyzing the single photon emission characteristics of NV centers embedded in an array of alumina-coated diamond nanopillars in high vacuum and low temperature (6K, high vacuum) environments under non-resonant (522 nm) laser exposure. We find that, in contrast to the oxygen-terminated diamond nanopillars, NV centers in the alumina-coated nanopillars demonstrate negligible change in the single photon purity and brightness over the course of laser exposure in vacuum. At low temperature, NV centers under alumina termination demonstrate stable single photon emission, whereas under oxygen termination the single photon purity degrades under high intensity laser exposure. Alumina surface passivation is therefore shown as a viable path toward the realization of robust NV-diamond based nanoscale sensing under non-ambient atmospheric environments, including using diamond scanning probes.

physics.app-ph↗

arXiv2026

When Majority Fails: Tight Bounds for Correlation Distillation Conjectures

We study two conjectures posed in the analysis of Boolean functions $f : \{-1, 1\}^n \to \{-1, 1\}$, in both of which, the Majority function plays a central role: the "Majority is Least Stable" (Benjamini et al., 1999) and the "Non-Interactive Correlation Distillation for Erasures" (Yang, 2004; O'Donnell and Wright, 2012). While both conjectures have been refuted in their originally stated form, we obtain a nearly tight characterization of the noise parameter regime in which each of the conjectures hold, for all $n \ge 5$. Whereas, for $n=3$, both conjectures hold in all noise parameter regimes. We state refined versions of both conjectures that we believe captures the spirit of the original conjectures.

cs.CC↗

arXiv2026

Selective braiding of different anyons in the even-denominator fractional quantum Hall effect

Even-denominator quantum Hall states can host several types of anyons with distinct exchange statistics. Depending on the anyon type, exchanging two quasiparticles can impart a phase to the many-body wave function or even transform it into a different state. Here, we realize a gate-tunable Fabry-Pérot interferometer with an embedded antidot that provides local control over the number of anyons within the interference loop. By independently tuning the magnetic field, carrier densities across the device, and the antidot potential, we access regimes in which localized anyons form reproducibly and measure the associated statistical phases $e^{i θ_\mathrm{braid}}$. We resolve braiding phases of $θ_{\mathrm{braid}}=π$ and $θ_{\mathrm{braid}}=\fracπ{2}$, which we attribute to $e/2$ quasiparticles encircling either $e/2$ or $e/4$ quasiparticles, respectively. We further observe switching between different anyon occupancies of the antidot over time, directly resolving individual anyon tunnelling events into the interference loop. Similar behavior occurs at filling factor one third. Our work addresses one of the two key challenges in observing non-Abelian braiding, which requires control of both localized and interfering anyon types.

cond-mat.mes-hall↗

arXiv2026

Denoising the US Census: Succinct Block Hierarchical Regression

The US Census Bureau Disclosure Avoidance System (DAS) balances confidentiality and utility requirements for the decennial US Census (Abowd et al., 2022). The DAS was used in the 2020 Census to produce demographic datasets critically used for legislative apportionment and redistricting, federal and state funding allocation, municipal and infrastructure planning, and scientific research. At the heart of DAS is TopDown, a heuristic post-processing method that combines billions of private noisy measurements across six geographic levels in order to produce new estimates that are consistent, more accurate, and satisfy certain structural constraints on the data. In this work, we introduce BlueDown, a new post-processing method that produces more accurate, consistent estimates while satisfying the same privacy guarantees and structural constraints. We obtain especially large accuracy improvements for aggregates at the county and tract levels on evaluation metrics proposed by the US Census Bureau. From a technical perspective, we develop a new algorithm for generalized least-squares regression that leverages the hierarchical structure of the measurements and that is statistically optimal among linear unbiased estimators. This reduces the computational dependence on the number of geographic regions measured from matrix multiplication time, which would be infeasible for census-scale data, to linear time. We incorporate the additional structural constraints by combining this regression algorithm with an optimization routine that extends TDA to support correlated measurements. We further improve the efficiency of our algorithm using succinct linear-algebraic operations that exploit symmetries in the structure of the measurements and constraints. We believe our hierarchical regression and succinct operations to be of independent interest.

cs.LG↗

arXiv2026

Interaction-driven quantum phase transitions between topological and crystalline orders of electrons

Topological and crystalline orders of electrons both benefit from enhanced Coulomb interactions in partially filled Landau levels. In bilayer graphene (BLG), the competition between fractional quantum Hall liquids and electronic crystals can be tuned electrostatically. Applying a displacement field leads to Landau-level crossings, where the interaction potential is strongly modified due to changes in the orbital wave functions. Here, we leverage this control to investigate phase transitions between topological and crystalline orders at constant filling factors in the lowest Landau level of BLG. Using transport measurements in high-quality hBN-encapsulated devices, we study transitions as a function of displacement field near crossings of $N=0$ and $N=1$ orbitals. The enhanced Landau-level mixing near the crossing stabilizes electronic crystals at all fractional fillings, including a resistive state at $ν= \frac{1}{3}$ and a reentrant integer quantum Hall state at $ν= \frac{7}{3}$. On the $N=0$ side, the activation energies of the crystal and fractional quantum Hall liquid vanish smoothly and symmetrically at the transition, while the $N=1$ transitions out of the crystal appear discontinuous. Additionally, we observe quantized plateaus forming near the crystal transition at half filling of the $N=0$ levels, suggesting a paired composite fermion state stabilized by Landau level mixing.

cond-mat.mes-hall↗

arXiv2026

Accelerating Scientific Research with Gemini: Case Studies and Common Techniques

Recent advances in large language models (LLMs) have opened new avenues for accelerating scientific research. While models are increasingly capable of assisting with routine tasks, their ability to contribute to novel, expert-level mathematical discovery is less understood. We present a collection of case studies demonstrating how researchers have successfully collaborated with advanced AI models, specifically Google's Gemini-based models (in particular Gemini Deep Think and its advanced variants), to solve open problems, refute conjectures, and generate new proofs across diverse areas in theoretical computer science, as well as other areas such as economics, optimization, and physics. Based on these experiences, we extract common techniques for effective human-AI collaboration in theoretical research, such as iterative refinement, problem decomposition, and cross-disciplinary knowledge transfer. While the majority of our results stem from this interactive, conversational methodology, we also highlight specific instances that push beyond standard chat interfaces. These include deploying the model as a rigorous adversarial reviewer to detect subtle flaws in existing proofs, and embedding it within a "neuro-symbolic" loop that autonomously writes and executes code to verify complex derivations. Together, these examples highlight the potential of AI not just as a tool for automation, but as a versatile, genuine partner in the creative process of scientific discovery.

cs.CL↗

arXiv2026

Computational Hardness of Private Coreset

We study the problem of differentially private (DP) computation of coreset for the $k$-means objective. For a given input set of points, a coreset is another set of points such that the $k$-means objective for any candidate solution is preserved up to a multiplicative $(1 \pm α)$ factor (and some additive factor). We prove the first computational lower bounds for this problem. Specifically, assuming the existence of one-way functions, we show that no polynomial-time $(ε, 1/n^{ω(1)})$-DP algorithm can compute a coreset for $k$-means in the $\ell_\infty$-metric for some constant $α> 0$ (and some constant additive factor), even for $k=3$. For $k$-means in the Euclidean metric, we show a similar result but only for $α= Θ\left(1/d^2\right)$, where $d$ is the dimension.

cs.CG↗

arXiv2026

Learning Multinomial Logits in $O(n \log n)$ time

A Multinomial Logit (MNL) model is composed of a finite universe of items $[n]=\{1,..., n\}$, each assigned a positive weight. A query specifies an admissible subset -- called a slate -- and the model chooses one item from that slate with probability proportional to its weight. This query model is also known as the Plackett-Luce model or conditional sampling oracle in the literature. Although MNLs have been studied extensively, a basic computational question remains open: given query access to slates, how efficiently can we learn weights so that, for every slate, the induced choice distribution is within total variation distance $\varepsilon$ of the ground truth? This question is central to MNL learning and has direct implications for modern recommender system interfaces. We provide two algorithms for this task, one with adaptive queries and one with non-adaptive queries. Each algorithm outputs an MNL $M'$ that induces, for each slate $S$, a distribution $M'_S$ on $S$ that is within $\varepsilon$ total variation distance of the true distribution. Our adaptive algorithm makes $O\left(\frac{n}{\varepsilon^{3}}\log n\right)$ queries, while our non-adaptive algorithm makes $O\left(\frac{n^{2}}{\varepsilon^{3}}\log n \log\frac{n}{\varepsilon}\right)$ queries. Both algorithms query only slates of size two and run in time proportional to their query complexity. We complement these upper bounds with lower bounds of $Ω\left(\frac{n}{\varepsilon^{2}}\log n\right)$ for adaptive queries and $Ω\left(\frac{n^{2}}{\varepsilon^{2}}\log n\right)$ for non-adaptive queries, thus proving that our adaptive algorithm is optimal in its dependence on the support size $n$, while the non-adaptive one is tight within a $\log n$ factor.

cs.DS↗

arXiv2025

Scaling Embedding Layers in Language Models

We propose $SCONE$ ($S$calable, $C$ontextualized, $O$ffloaded, $N$-gram $E$mbedding), a new method for extending input embedding layers to enhance language model performance. To avoid increased decoding costs, $SCONE$ retains the original vocabulary while introducing embeddings for a set of frequent n-grams. These embeddings provide contextualized representation for each input token and are learned with a separate model during training. After training, embeddings are precomputed and stored in off-accelerator memory; during inference, querying them has minimal impact on latency due to the low complexity of embedding lookups. $SCONE$ enables two new scaling strategies: increasing the number of n-gram embeddings and scaling the model used to learn them, both while maintaining fixed accelerator usage during inference (in terms of FLOPS and memory). We show that scaling both aspects enables a model with 1B accelerator-resident parameters to outperform a 1.9B-parameter baseline across diverse corpora, while using only about half the FLOPS and accelerator memory during inference.

cs.CL↗

arXiv2025

VaultGemma: A Differentially Private Gemma Model

We introduce VaultGemma 1B, a 1 billion parameter model within the Gemma family, fully trained with differential privacy. Pretrained on the identical data mixture used for the Gemma 2 series, VaultGemma 1B represents a significant step forward in privacy-preserving large language models. We openly release this model to the community

cs.CR↗

arXiv2025

Probing transition rates, nuclear moments and electric dipole polarizability in nobelium using multireference FSRCC and PRCC theories

We employ an all-particle multireference Fock-space relativistic coupled-cluster (FSRCC) theory to compute the ionization potential, excitation energy, transition rate and hyperfine structure constants associated with $7s^2\;^{1}S_{0}\rightarrow 7s7p\;^{3}P_{1}$ and $7s^2\;^{1}S_{0}\rightarrow 7s7p\;^1P_{1}$ transitions in nobelium (No). Using our state-of-the-art calculations in conjunction with available experimental data \cite{raeder-18}, we extract the values of nuclear magnetic dipole ($μ$) and electric quadrupole ($Q$) moments for $^{253}$No. Further, information on nuclear deformation in even-mass isotopes is extracted from the isotope shift calculations. Moreover, we employ a perturbed relativistic coupled-cluster (PRCC) theory to compute the ground state electric dipole polarizability of No. In addition, to assess the accuracy of our calculations, we compute the ionization potential and dipole polarizability of lighter homolog ytterbium (Yb). To account for strong relativistic and quantum electrodynamical (QED) effects in No, we incorporate the corrections from Breit interaction, vacuum polarization and self-energy in our calculations. The contributions from triple excitations in coupled-cluster is accounted perturbatively. Our calculations reveal a significant contribution of $\approx$10\% from the perturbative triples to the transition rate of $7s^2\;^1S_{0}\rightarrow 7s7p\;^3P_{1}$ transition. The largest cumulative contribution from Breit+QED is observed to be $\approx$4\%, to the magnetic dipole hyperfine structure constant of $7s7p\;^1P_{1}$ state. Our study provides a comprehensive understanding of atomic and nuclear properties of nobelium with valuable insights into the electron correlation and relativistic effects in superheavy elements.

physics.atom-ph↗

arXiv2025

Impact of irradiation conditions on the magnetic field sensitivity of spin defects in hBN nano flakes

We study $V_{\mathrm{B}}^-$ centres generated by helium focused ion beam (FIB) irradiation in thin ($\sim$70 nm) hBN nanoflakes, in order to investigate the effect of implantation conditions on the key parameters that influence the magnetic field sensitivity of $V_{\mathrm{B}}^-$ quantum sensors. Using a combination of photoluminescence, optically detected magnetic resonance, and Raman spectroscopy, we examine the competing factors of maximising signal intensity through larger $V_{\mathrm{B}}^-$ concentration against the degradation in spin coherence and lattice quality observed at high ion fluences. Our results indicate that both the $V_{\mathrm{B}}^-$ spin properties and hBN lattice parameters are largely preserved up to an ion fluence of $10^{14}$ ions/cm$^2$, and beyond this significant degradation occurs in both. At the optimal implantation dose, an AC magnetic sensitivity of $\sim 1\,μ\mathrm{T}/\sqrt{\mathrm{Hz}}$ is achieved. Using the patterned implantation enabled by the FIB, we find that $V_{\mathrm{B}}^-$ centres and the associated lattice damage are well localised to the implanted regions. This work demonstrates how careful selection of fabrication parameters can be used to optimise the properties of $V_{\mathrm{B}}^-$ centres in hBN, supporting their application as quantum sensors based on 2D materials.

physics.app-ph↗

arXiv2025

Likelihood-free inference for gravitational-wave data analysis and public alerts

Rapid and reliable detection and dissemination of source parameter estimation data products from gravitational-wave events, especially sky localization, is critical for maximizing the potential of multi-messenger astronomy. Machine learning based detection and parameter estimation algorithms are emerging as production ready alternatives to traditional approaches. Here, we report validation studies of AMPLFI, a likelihood-free inference solution to low-latency parameter estimation of binary black holes. We use simulated signals added into data from the LIGO-Virgo-KAGRA's (LVK's) third observing run (O3) to compare sky localization performance with BAYESTAR, the algorithm currently in production for rapid sky localization of candidates from matched-filter pipelines. We demonstrate sky localization performance, measured by searched area and volume, to be equivalent with BAYESTAR. We show accurate reconstruction of source parameters with uncertainties for use distributing low-latency coarse-grained chirp mass information. In addition, we analyze several candidate events reported by the LVK in the third gravitational-wave transient catalog (GWTC-3) and show consistency with the LVK's analysis. Altogether, we demonstrate AMPLFI's ability to produce data products for low-latency public alerts.

gr-qc↗

arXiv2025

Urania: Differentially Private Insights into AI Use

We introduce $Urania$, a novel framework for generating insights about LLM chatbot interactions with rigorous differential privacy (DP) guarantees. The framework employs a private clustering mechanism and innovative keyword extraction methods, including frequency-based, TF-IDF-based, and LLM-guided approaches. By leveraging DP tools such as clustering, partition selection, and histogram-based summarization, $Urania$ provides end-to-end privacy protection. Our evaluation assesses lexical and semantic content preservation, pair similarity, and LLM-based metrics, benchmarking against a non-private Clio-inspired pipeline (Tamkin et al., 2024). Moreover, we develop a simple empirical privacy evaluation that demonstrates the enhanced robustness of our DP pipeline. The results show the framework's ability to extract meaningful conversational insights while maintaining stringent user privacy, effectively balancing data utility with privacy preservation.

cs.LG↗

arXiv2025

Private Hyperparameter Tuning with Ex-Post Guarantee

The conventional approach in differential privacy (DP) literature formulates the privacy-utility trade-off with a "privacy-first" perspective: for a predetermined level of privacy, a certain utility is achievable. However, practitioners often operate under a "utility-first" paradigm, prioritizing a desired level of utility and then determining the corresponding privacy cost. Wu et al. [2019] initiated a formal study of this "utility-first" perspective by introducing ex-post DP. They demonstrated that by adding correlated Laplace noise and progressively reducing it on demand, a sequence of increasingly accurate estimates of a private parameter can be generated, with the privacy cost attributed only to the least noisy iterate released. This led to a Laplace mechanism variant that achieves a specified utility with minimal privacy loss. However, their work, and similar findings by Whitehouse et al. [2022], are primarily limited to simple mechanisms based on Laplace or Gaussian noise. In this paper, we significantly generalize these results. In particular, we extend the work of Wu et al. [2019] and Liu and Talwar [2019] to support any sequence of private estimators, incurring at most a doubling of the original privacy budget. Furthermore, we demonstrate that hyperparameter tuning for these estimators, including the selection of an optimal privacy budget, can be performed without additional privacy cost. Finally, we extend our results to ex-post Renyi DP, further broadening the applicability of utility-first privacy mechanisms.

cs.CR↗