arXiv subjects

Girish Kumar

Publications and source records attributed to Girish Kumar.

At least 19 recordsLinked to original sources

arXiv2026

Axion-like particles and sterile neutrinos solve the $B\to K\nu\bar\nu$ and $B\to \pi K$ puzzles

The recent measurement of the branching ratio of $B^+ \to K^+ + \mathrm{inv}$ (where ``inv'' denotes invisible states) by the Belle II collaboration is enhanced relative to the standard model expectation by 2.7$\sigma$. An older puzzle persists in measurements of the branching ratios and CP asymmetries of $B \to \pi K $ decays. We address these two anomalies in flavor-changing neutral current $B$ decays, with a short-lived axion-like particle (ALP) with mass close to that of the $\pi^0$. In the model with the minimum number of new couplings, the ALP has couplings to the photon, top quark and a heavy sterile neutrino. The ALP contributes to the $B \to \pi^0 K $ decays by mixing with the $\pi^0$. It contributes to $B^+ \to K^+ + \mathrm{inv}$ by its off-shell coupling to sterile neutrino pairs. The model can explain the excess in the total rate, but not the observed distribution of signal events. We make predictions for all $B \to K^{(*)} + \mathrm{inv}$ modes and for the rare kaon decays, $K^+ \to \pi^+ + \mathrm{inv}$ and $K_L \to \pi^0 + \mathrm{inv}$. We find an appreciable contribution to the magnetic moment of the muon, and negligible contributions to the magnetic moment of the electron and $b \to s e^+ e^-$.

hep-ph↗

arXiv2026

Axion-like Particles and Lepton Flavor Violation in Muonic Atoms

We explore the potential of the Mu2e experiment to probe the lepton-flavor-violating process $\mu^- e^- \to e^- e^-$ in a muonic atom within a simplified axion-like particle (ALP) framework featuring flavor-violating $e$-$\mu$ couplings and a flavor-diagonal pseudoscalar coupling to electrons, which also allows for possible invisible ALP decays into a dark sector. We compute the ALP-mediated contribution to the transition rate and show that, at fixed couplings, the branching ratio increases for lighter mediators and scales as $(Z-1)^3$, favoring heavier nuclei. We compare the model against constraints from $\mu\to e\gamma$, $\mu\to 3e$, $\mu\to e\gamma\gamma$, $\mu\to e+\mathrm{inv}$, and muonium-antimuonium conversion, as well as from the anomalous magnetic moments of the electron and muon. Additional astrophysical and beam-dump limits on the electron coupling are also discussed. A key result is that $\Delta a_e$ provides one of the most stringent probes of the parameter space and, in the global scan, excludes the largest fraction of sampled points. After applying the laboratory constraints used in the scan, the viable branching ratio for $\mu^- e^- \to e^- e^-$ in aluminum drops to at most $\mathcal{O}(10^{-20})$, while the resonant region $2m_e<m_a<m_\mu-m_e$ is much more heavily suppressed. The highest achievable values are closely tied to $\mathcal{B}(\mu\to 3e)$ near its current limit, indicating that the upcoming Mu3e experiment will explore the most promising region relevant for this muonic-atom signal. Our analysis shows that, although a light ALP can parametrically enhance $\mu^- e^- \to e^- e^-$ at fixed couplings, existing bounds -- especially $\Delta a_e$, $\mu\to 3e$, $\mu\to e\gamma$, and muonium oscillations -- severely limit the observable rate.

hep-ph↗

arXiv2026

WHBench: Evaluating Frontier LLMs with Expert-in-the-Loop Validation on Women's Health Topics

Large language models are increasingly used for medical guidance, but women's health remains under-evaluated in benchmark design. We present the Women's Health Benchmark (WHBench), a targeted evaluation suite of 47 expert-crafted scenarios across 10 women's health topics, designed to expose clinically meaningful failure modes including outdated guidelines, unsafe omissions, dosing errors, and equity-related blind spots. We evaluate 22 models using a 23-criterion rubric spanning clinical accuracy, completeness, safety, communication quality, instruction following, equity, uncertainty handling, and guideline adherence, with safety-weighted penalties and server-side score recalculation. Across 3,102 attempted responses (3,100 scored), no model mean performance exceeds 75 percent; the best model reaches 72.1 percent. Even top models show low fully correct rates and substantial variation in harm rates. Inter-rater reliability is moderate at the response label level but high for model ranking, supporting WHBench utility for comparative system evaluation while highlighting the need for expert oversight in clinical deployment. WHBench provides a public, failure-mode-aware benchmark to track safer and more equitable progress in womens health AI.

cs.CL↗

arXiv2025

High-Fidelity Synthetic ECG Generation via Mel-Spectrogram Informed Diffusion Training

The development of machine learning for cardiac care is severely hampered by privacy restrictions on sharing real patient electrocardiogram (ECG) data. Although generative AI offers a promising solution, the real-world use of existing model-synthesized ECGs is limited by persistent gaps in trustworthiness and clinical utility. In this work, we address two major shortcomings of current generative ECG methods: insufficient morphological fidelity and the inability to generate personalized, patient-specific physiological signals. To address these gaps, we build on a conditional diffusion-based Structured State Space Model (SSSD-ECG) with two principled innovations: (1) MIDT-ECG (Mel-Spectrogram Informed Diffusion Training), a novel training paradigm with time-frequency domain supervision to enforce physiological structural realism, and (2) multi-modal demographic conditioning to enable patient-specific synthesis. We comprehensively evaluate our approach on the PTB-XL dataset, assessing the synthesized ECG signals on fidelity, clinical coherence, privacy preservation, and downstream task utility. MIDT-ECG achieves substantial gains: it improves morphological coherence, preserves strong privacy guarantees with all metrics evaluated exceeding the baseline by 4-8%, and notably reduces the interlead correlation error by an average of 74%, while demographic conditioning enhances signal-to-noise ratio and personalization. In critical low-data regimes, a classifier trained on datasets supplemented with our synthetic ECGs achieves performance comparable to a classifier trained solely on real data. Together, we demonstrate that ECG synthesizers, trained with the proposed time-frequency structural regularization scheme, can serve as personalized, high-fidelity, privacy-preserving surrogates when real data are scarce, advancing the responsible use of generative AI in healthcare.

cs.LG↗

arXiv2024

Finite width effects in nonleptonic D-decays

Many analyses of two-body non-leptonic decays of $D$-mesons rely on flavor SU(3) symmetry relations and fits of experimental data of decays rates to extract the universal transition amplitudes. Such fits assume that the final state mesons are well-defined asymptotic states of QCD. We develop a technique to take into account the finite width effects of the final state mesons and study their effects on the extracted values of transition amplitudes.

hep-ph↗

arXiv2024

Differentially Private Synthetic High-dimensional Tabular Stream

While differentially private synthetic data generation has been explored extensively in the literature, how to update this data in the future if the underlying private data changes is much less understood. We propose an algorithmic framework for streaming data that generates multiple synthetic datasets over time, tracking changes in the underlying private data. Our algorithm satisfies differential privacy for the entire input stream (continual differential privacy) and can be used for high-dimensional tabular data. Furthermore, we show the utility of our method via experiments on real-world datasets. The proposed algorithm builds upon a popular select, measure, fit, and iterate paradigm (used by offline synthetic data generation algorithms) and private counters for streams.

cs.CR↗

arXiv2024

Strategic Pseudo-Goal Perturbation for Deadlock-Free Multi-Agent Navigation in Social Mini-Games

This work introduces a Strategic Pseudo-Goal Perturbation (SPGP) technique, a novel approach to resolve deadlock situations in multi-agent navigation scenarios. Leveraging the robust framework of Safety Barrier Certificates, our method integrates a strategic perturbation mechanism that guides agents through social mini-games where deadlock and collision occur frequently. The method adopts a strategic calculation process where agents, upon encountering a deadlock select a pseudo goal within a predefined radius around the current position to resolve the deadlock among agents. The calculation is based on controlled strategic algorithm, ensuring that deviation towards pseudo-goal is both purposeful and effective in resolution of deadlock. Once the agent reaches the pseudo goal, it resumes the path towards the original goal, thereby enhancing navigational efficiency and safety. Experimental results demonstrates SPGP's efficacy in reducing deadlock instances and improving overall system throughput in variety of multi-agent navigation scenarios.

cs.MA↗

arXiv2024

Constraints on light dark sector particles from lifetime difference of heavy neutral mesons

Heavy meson decays with missing energy in the final state offer interesting avenues to search for light invisible new physics such as dark matter (DM). In this context, we show that such new physics (NP) interactions also affect lifetime difference in neutral meson-antimeson mixing. We consider general dimension-six effective quark interactions involving a pair of DM particles and calculate their contributions to lifetime difference in beauty and charm meson systems. We use the latest data on mixing observables to constrain the relevant effective operators. We find that lifetime differences provide novel and complementary flavor constraints compared to those obtained from heavy meson decays.

hep-ph↗

arXiv2024

An Algorithm for Streaming Differentially Private Data

Much of the research in differential privacy has focused on offline applications with the assumption that all data is available at once. When these algorithms are applied in practice to streams where data is collected over time, this either violates the privacy guarantees or results in poor utility. We derive an algorithm for differentially private synthetic streaming data generation, especially curated towards spatial datasets. Furthermore, we provide a general framework for online selective counting among a collection of queries which forms a basis for many tasks such as query answering and synthetic data generation. The utility of our algorithm is verified on both real-world and simulated datasets.

cs.DB↗

arXiv2023

Discovery Prospects for Electron and Neutron Electric Dipole Moments in the General Two Higgs Doublet Model

Baryon asymmetry of the Universe offers one of the strongest hints for physics Beyond the Standard Model (BSM). Remarkably, in the general two Higgs Doublet Model (g2HDM) that possesses a second set of Yukawa matrices, one can have electroweak baryogenesis (EWBG) while the electron electric dipole moment (eEDM) is evaded by a natural flavor tuning that echoes SM. We show that eEDM may first emerge around $10^{-30}\,e$ cm or so, followed by neutron EDM (nEDM) down to $10^{-27}\,e$ cm. We illustrate a cancellation mechanism for nEDM itself, which in turn can be probed when a facility capable of pushing down to $10^{-28}\,e$ cm becomes available.

hep-ph↗

arXiv2023

Probing Baryogenesis with Radiative Beauty Decay and Electron EDM

With the Large Hadron Collider (LHC) running, we should probe electroweak baryogenesis (EWBG) while probing $CP$ violation (CPV) with electron electric dipole moment (eEDM). Rooted in the flavor structure of the Standard Model (SM), the general two Higgs doublet model (g2HDM) with a second set of Yukawa couplings can deliver EWBG while surviving eEDM. We point out a chiral-enhanced top-bottom interference effect that makes $b \to s\gamma$ decay an exquisite window on EWBG and eEDM, and illustrate the importance of the $\Delta A_{CP}$ observable at Belle II.

hep-ph↗

arXiv2022

Interplay of the charged Higgs effects in $R_{D^{(\ast)}}$, $b\to s \ell^+\ell^-$ and $W$-mass

Current data on semileptonic charged- and neutral-current $B$ decays show deviations from the predictions of the Standard Model. It is well known that a charged Higgs boson, belonging to the two-Higgs doublet model without $Z_2$ symmetry, offers one of the simplest solution to the charged-current $B$ decays. We show that this solution naturally induces a negative shift of $\mathcal{O}(1)$ in the Wilson coefficient ($C_{9\ell}$) of operator $(\bar s_L\gamma_\mu b_L)(\bar \ell\gamma^\mu \ell)$, potentially resolving the tension in neutral-current $B$ decays as well. Interestingly, the lepton universality ratios in $b\to s \ell^+\ell^-$ decays, in tune with the recent LHCb result, remain SM-like. Precision constraints from neutral $B$ and $K$ meson mixing, decays $B_c\to \tau\bar\nu$, $B\to X_s\gamma$, and leptonic decays of $\tau$ and $Z$ can be satisfied. Furthermore, a positive shift in $W$-boson mass, nicely in agreement with the CDF measurement, is also possible, requiring the neutral scalars to be heavier than the charged Higgs but within the sub-TeV region.

hep-ph↗

arXiv2022

$\Lambda_b\rightarrow \Lambda(\to p \pi^-) \ell^+\ell^-$ as probe of CP-violating New Physics

We investigate the possible sizes of all the CP-violating asymmetries offered by the angular distribution of rare decay $\Lambda_b\rightarrow \Lambda(\to p \pi^-) \ell^+\ell^-$ in the Standard Model and new physics scenarios motivated by the recent $b\to s \ell^+\ell^-$ anomalies. We work in a model-independent effective theory framework and discuss the sensitivity of CP asymmetries to new ${O}_{9,10}$ operators and their chirality flipped counterparts. We find that the size of many of the CP asymmetries can be at the level of a few percent in new physics scenarios consistent with current $b\to s\ell^+\ell^-$ data at a level of $1\sigma$. We emphasize that measurements of these CP asymmetries can be used to discriminate different new physics scenarios in $b\to s \ell^+\ell^-$.

hep-ph↗

arXiv2022

Kaon processes in general 2HDM

We discuss new physics (NP) contributions to kaon mixing parameter $\varepsilon_K$, direct CP violation parameter $\varepsilon^\prime/\varepsilon$ of $K\to \pi\pi$, and rare decays $K^+\to \pi^+\nu\bar\nu$, $K_L\to \pi^0\nu\bar\nu$ and $K_{L, S}\to \mu^+ \mu^- $ in the context of general two Higgs doublet model. We focus on contributions of top quark related exotic couplings, and show that simultaneous presence of flavor conserving and flavor violating interactions can lead to large NP effects in kaon sector, while being consistent with the stringent constraints from B physics observables such as $B_{s(d)}$-$\bar B_{s(d)}$ mixing, $B_s\to \mu^+ \mu^-$, and $b\to s\gamma$. We stress on the importance of correlations between $\varepsilon_K$, $K^+\to \pi^+\nu\bar\nu$ and $B_s\to \mu^+ \mu^-$ that can be exploited to distinguish the parameter space corresponding to a light (sub-TeV) or heavy (TeV) scale charged Higgs boson.

hep-ph↗

arXiv2022

Synthetic Text Generation with Differential Privacy: A Simple and Practical Recipe

Privacy concerns have attracted increasing attention in data-driven products due to the tendency of machine learning models to memorize sensitive training data. Generating synthetic versions of such data with a formal privacy guarantee, such as differential privacy (DP), provides a promising path to mitigating these privacy concerns, but previous approaches in this direction have typically failed to produce synthetic data of high quality. In this work, we show that a simple and practical recipe in the text domain is effective: simply fine-tuning a pretrained generative language model with DP enables the model to generate useful synthetic text with strong privacy protection. Through extensive empirical analyses on both benchmark and private customer data, we demonstrate that our method produces synthetic text that is competitive in terms of utility with its non-private counterpart, meanwhile providing strong protection against potential privacy leakages.

cs.CL↗

arXiv2022

Enhanced $B_q \to \ell\ell^\prime$ and $B \to (K, \pi) \ell\ell^\prime$ in light of $(g-2)_\mu$

We study lepton flavor violating (LFV) $B$ decays in a general two Higgs doublet model with sub-TeV exotic scalars. Two different parameter spaces are explored: one dominated by extra top Yukawa coupling $\rho_{tt}$, the other by LFV couplings relevant for the muon $g-2$ anomaly. In the first case, flavor constraints such as $\ell \to \ell^\prime\gamma$, $h \to \ell\ell^\prime$ imply LFV $B$ decays are far below experimental sensitivities. The second case needs to be close to the alignment limit, but $B_q \to \tau\mu$ and $B \to (K, \pi) \tau\mu$ rates can lie within the sensitivities of Belle II and LHCb Upgrade II. Neutral $B$ meson mixings and $B,\, K,\, \pi \to \ell \nu$ decays provide important flavor constraints on parameter space. $B$ decays involving $e$-$\tau$ violation are constrained by $\mu\to e$ processes.

hep-ph↗

arXiv2022

Strange processes in general two Higgs doublet model

In the general two Higgs doublet model (g2HDM) that has extra Yukawa couplings, we analyze their New Physics (NP) contributions to kaon mixing parameter $\varepsilon_K$, direct CP violation parameter $\varepsilon^\prime/\varepsilon$, and rare $K^+ \to \pi^+\nu\bar\nu$, $K_L \to \pi^0\nu\bar\nu$, and $K_{L, S} \to \mu^+ \mu^-$ decays. We study correlations between these observables, and call special attention to a unique complementarity of kaon mixing and rare $K\to \pi \nu\bar\nu$ decays in probing the exotic Higgs mass spectrum of g2HDM. The importance of kaon physics in probing NP vis-\`a-vis B physics is stressed. One unexpected feature we uncover is the special sensitivity of $K^+ \to \pi^+\nu\bar\nu$ to TeV scale charged Higgs boson.

hep-ph↗

arXiv2021

Charged Lepton EDM with Extra Yukawa Couplings

In a two Higgs doublet model with extra Yukawa couplings, we assess the new physics contributions to charged lepton electric dipole moment. We focus especially on muon (and tau) EDM where in the coming decade several experiments $-$ Muon g-2, J-PARC and PSI (and Belle II) $-$ will push sensitivities down by several orders of magnitude. With the working assumption that extra Yukawa couplings are analogous to SM ones in strength and taking exotic scalar masses in sub-TeV range, we find that $\mu$ and $\tau$ EDM can be enhanced to values larger than new physics scenarios that scale with lepton mass. The main effect comes from the flavor-conserving extra top coupling $\rho_{tt}$ via well-known two-loop diagrams. Deviating from our working assumption, if the muon $g-2$ anomaly arises from the one-loop diagram, driven by singly enhanced lepton flavor violating $\rho_{\tau\mu}$ coupling, it can also induce rather large muon EDM, accessible at upcoming experiments.

hep-ph↗