SearcharxivSearch

arXiv subjects

Jay Desai

Publications and source records attributed to Jay Desai.

14 recordsLinked to original sources

USMEFT as a tool for discovery of universal new physics at high luminosity LHC

We analyze the potential of the Universal SMEFT as a tool to study universal new physics in the neutral and charged Drell-Yan processes at the high-luminosity LHC. In order to do so, we generated pseudo-data containing the contributions of beyond the standard model vector bosons to the dilepton production. Our results show that fits with the minimal theoretical bias can be used not only to unveil the existence of new physics but also to accurately extract its properties. Moreover, the results are rather stable with respect to the order of truncation of the effective field theory.

hep-ph

Positivity in Massive Spin-3/2 EFTs and the Planck-Suppressed Neighbourhood of Supergravity

It is well known that a strictly massless spin-$3/2$ particle can interact consistently only within supergravity. Recently, positivity arguments have shown that an effective field theory of a massive Majorana spin-$3/2$ particle admits a smooth $m \to 0$ limit only if a graviton is present and the four-fermion contact interactions are tuned to the values dictated by $\mathcal{N}=1$ supergravity. In this work, we investigate how this limit is approached at finite mass. Assuming that the graviton $t$-channel pole can be discarded, we derive non-forward, tree-level dispersive bounds on massive spin-3/2 contact operators and determine the region of effective couplings consistent with unitarity and analyticity. For sufficiently small $m$, we find that the allowed parameter space forms a bounded, Planck-suppressed neighbourhood of the supergravity point, defined by the supergravity values of the four-fermion couplings. The supergravity point lies on the boundary of this region. In the regime $m \ll M_{\rm Pl}$, the volume of the allowed region scales parametrically as \[ \mathrm{Vol} \sim \frac{m^{6}}{M_{\rm Pl}^{6}} \, , \] and shrinks to zero as $m \to 0$, smoothly reproducing the massless-limit results. The allowed region becomes unbounded when mass approaches the Planck scale. We further analyze the effect of including additional light scalar and pseudo-scalar degrees of freedom, motivated by the Polonyi model, and find that their couplings are also bounded in a way similar to the contact couplings and that it doesn't enlarge the allowed contact coupling space.

hep-th

Stable Causality and Microcausality for Drummond-Hathrell Photons

Local superluminal photon propagation arises at $\mathcal{O}(\alpha/m_e^2)$ in the Drummond Hathrell (DH) effective action obtained by integrating out the electron in QED coupled to gravity. Whether such superluminality implies a genuine violation of causality in curved spacetime is subtle and remains conceptually nontrivial. In this work we revisit this question using two complementary and largely symmetry-independent diagnostics. First, we analyse the global causal structure of the effective (optical) metric governing DH photon propagation and identify conditions under which it remains stably causal, thereby excluding the formation of closed causal curves. Second, from a quantum-field-theoretic perspective, we examine microcausality by treating the gravitational background as a fixed Lorentz-breaking field and applying flat-spacetime analyticity bounds to the photon commutator within the geometric-optics regime of the EFT. For two representative examples, a circular photon orbit in Schwarzschild and a linear trajectory in a two-black-hole geometry, we find that, within the regime of validity of the DH effective theory, both diagnostics indicate that the superluminal photon propagation is causally benign. Our results do not constitute a general definition of microcausality in curved spacetime, but provide a controlled and instructive check of causal consistency for EFT superluminality in gravitational backgrounds.

hep-th

Positivity at 1-Loop: Bounds on Photon and Gluon EFTs

In this paper, we attempt to derive ``positivity" bounds on Photon and Gluon Effective Field Theories (EFTs) at one loop level. While for the Photon case, the one loop amplitude is IR finite and well defined in the forward limit, earlier studies failed to obtain a dispersive bound on dimension-12 operators due to the dependence of the ``arc integral" on the artificial low-energy scale. We show that this awkward dependence can be taken care of by analysing the ultra-violet (UV) side of dispersion relation closely. In particular, we derive an IR safe and RG improved bound at 1-loop. Thereafter, we perform a similar analysis on the Gluon EFT, which has additional complications due to ill-defined forward limit and IR divergences at 1-loop. We show that even in this case, one can get a meaningful bound at 1-loop.

hep-ph

Drell-Yan production in universal theories beyond dimension-six SMEFT

We study Drell-Yan production in universal theories consistently including effects beyond dimension six in the SMEFT. Within universal SMEFT and with $C$ and $P$ conservation we find that eleven dimension-eight operators contribute in addition to the six contributing at dimension-six. We first work in an operator basis in which operators with higher derivatives of the bosonic fields have been rotated by equations of motion in favor of combinations of operators involving SM fermion currents. We derive the general form of the amplitudes consistently in the expansion to ${\cal O}(\Lambda^{-4})$ and identify eight combinations of the 17 Wilson coefficients which are physically distinguishable by studying the invariant mass distribution of the lepton pairs produced. We then introduce an extension of the parametrization of universal effects in terms of oblique parameters obtained by linearly expanding the self-energies of the electroweak gauge-bosons to ${\cal O}(q^6)$. It contains eleven oblique parameters of which only eight are generated within SMEFT at dimension-eight: $\widehat{S}$, $\widehat{T}$, $W$, $Y$, $\widehat{U}$, $X$, plus two additional which we label $W'$ and $Y'$ and show how they match at linear order with the eight identified combinations of operator coefficients. We then perform a combined analysis of a variety of LHC data on the neutral- and charged-current Drell-Yan processes with the aim of constraining the eight combinations. We compare and combine the LHC bounds with those from electroweak precision $W$ and $Z$ pole observables which can only provide constraints in four directions of the eight-parameter space. We present the results in terms of limits on the eight effective Wilson coefficients as well as on the eight oblique parameters. In each case, we study the dependence of the derived constraints on the order of the expansion considered.

hep-ph

HR-Agent: A Task-Oriented Dialogue (TOD) LLM Agent Tailored for HR Applications

Recent LLM (Large Language Models) advancements benefit many fields such as education and finance, but HR has hundreds of repetitive processes, such as access requests, medical claim filing and time-off submissions, which are unaddressed. We relate these tasks to the LLM agent, which has addressed tasks such as writing assisting and customer support. We present HR-Agent, an efficient, confidential, and HR-specific LLM-based task-oriented dialogue system tailored for automating repetitive HR processes such as medical claims and access requests. Since conversation data is not sent to an LLM during inference, it preserves confidentiality required in HR-related tasks.

cs.CL

LOLAMEME: Logic, Language, Memory, Mechanistic Framework

The performance of Large Language Models has achieved superhuman breadth with unprecedented depth. At the same time, the language models are mostly black box models and the underlying mechanisms for performance have been evaluated using synthetic or mechanistic schemes. We extend current mechanistic schemes to incorporate Logic, memory, and nuances of Language such as latent structure. The proposed framework is called LOLAMEME and we provide two instantiations of LOLAMEME: LoLa and MeMe languages. We then consider two generative language model architectures: transformer-based GPT-2 and convolution-based Hyena. We propose the hybrid architecture T HEX and use LOLAMEME framework is used to compare three architectures. T HEX outperforms GPT-2 and Hyena on select tasks.

cs.LG

JADS: A Framework for Self-supervised Joint Aspect Discovery and Summarization

To generate summaries that include multiple aspects or topics for text documents, most approaches use clustering or topic modeling to group relevant sentences and then generate a summary for each group. These approaches struggle to optimize the summarization and clustering algorithms jointly. On the other hand, aspect-based summarization requires known aspects. Our solution integrates topic discovery and summarization into a single step. Given text data, our Joint Aspect Discovery and Summarization algorithm (JADS) discovers aspects from the input and generates a summary of the topics, in one step. We propose a self-supervised framework that creates a labeled dataset by first mixing sentences from multiple documents (e.g., CNN/DailyMail articles) as the input and then uses the article summaries from the mixture as the labels. The JADS model outperforms the two-step baselines. With pretraining, the model achieves better performance and stability. Furthermore, embeddings derived from JADS exhibit superior clustering capabilities. Our proposed method achieves higher semantic alignment with ground truth and is factual.

cs.AI

Dimension-eight Operator Basis for Universal Standard Model Effective Field Theory

We present the basis of dimension-eight operators associated with universal theories. We first derive a complete list of independent dimension-eight operators formed with the Standard Model bosonic fields characteristic of such universal new physics scenarios. Without imposing C or P symmetries the basis contains 175 operators -- that is, the assumption of Universality reduces the number of independent SMEFT coefficients at dimension eight from 44807 to 175. 89 of the 175 universal operators are included in the general dimension-eight operator basis in the literature. The 86 additional operators involve higher derivatives of the Standard Model bosonic fields and can be rotated in favor of operators involving fermions using the Standard Model equations of motion for the bosonic fields. By doing so we obtain the allowed fermionic operators generated in this class of models which we map into the corresponding 86 independent combinations of operators in the dimension-eight basis of Ref. arXiv:2005.00059

hep-ph

KDSTM: Neural Semi-supervised Topic Modeling with Knowledge Distillation

In text classification tasks, fine tuning pretrained language models like BERT and GPT-3 yields competitive accuracy; however, both methods require pretraining on large text datasets. In contrast, general topic modeling methods possess the advantage of analyzing documents to extract meaningful patterns of words without the need of pretraining. To leverage topic modeling's unsupervised insights extraction on text classification tasks, we develop the Knowledge Distillation Semi-supervised Topic Modeling (KDSTM). KDSTM requires no pretrained embeddings, few labeled documents and is efficient to train, making it ideal under resource constrained settings. Across a variety of datasets, our method outperforms existing supervised topic modeling methods in classification accuracy, robustness and efficiency and achieves similar performance compare to state of the art weakly supervised text classification methods.

cs.CL

S2vNTM: Semi-supervised vMF Neural Topic Modeling

Language model based methods are powerful techniques for text classification. However, the models have several shortcomings. (1) It is difficult to integrate human knowledge such as keywords. (2) It needs a lot of resources to train the models. (3) It relied on large text data to pretrain. In this paper, we propose Semi-Supervised vMF Neural Topic Modeling (S2vNTM) to overcome these difficulties. S2vNTM takes a few seed keywords as input for topics. S2vNTM leverages the pattern of keywords to identify potential topics, as well as optimize the quality of topics' keywords sets. Across a variety of datasets, S2vNTM outperforms existing semi-supervised topic modeling methods in classification accuracy with limited keywords provided. S2vNTM is at least twice as fast as baselines.

cs.CL

Two-Stage Violence Detection Using ViTPose and Classification Models at Smart Airports

This study introduces an innovative violence detection framework tailored to the unique requirements of smart airports, where prompt responses to violent situations are crucial. The proposed framework harnesses the power of ViTPose for human pose estimation. It employs a CNN - BiLSTM network to analyse spatial and temporal information within keypoints sequences, enabling the accurate classification of violent behaviour in real time. Seamlessly integrated within the SAFE (Situational Awareness for Enhanced Security framework of SAAB, the solution underwent integrated testing to ensure robust performance in real world scenarios. The AIRTLab dataset, characterized by its high video quality and relevance to surveillance scenarios, is utilized in this study to enhance the model's accuracy and mitigate false positives. As airports face increased foot traffic in the post pandemic era, implementing AI driven violence detection systems, such as the one proposed, is paramount for improving security, expediting response times, and promoting data informed decision making. The implementation of this framework not only diminishes the probability of violent events but also assists surveillance teams in effectively addressing potential threats, ultimately fostering a more secure and protected aviation sector. Codes are available at: https://github.com/Asami-1/GDP.

cs.CV

Impact of dimension-eight SMEFT operators in the EWPO and Triple Gauge Couplings analysis in Universal SMEFT

We perform a complete study of the electroweak precision observables and electroweak gauge boson pair production in terms of the SMEFT up to ${\cal O}(1/Λ^4)$ under the assumption of universal, C and P conserving new physics. We show that the analysis of data from those two sectors allows us to obtain closed constraints in the relevant parameter space in this scenario. In particular we find that the Large Hadron Collider data can independently constrain the Wilson coefficients of the dimension-six and -eight operators directly contributing to the triple gauge boson vertices. Our results show that the impact of dimension-eight operators in the study of triple gauge couplings is small.

hep-ph

Attention-based Region of Interest (ROI) Detection for Speech Emotion Recognition

Automatic emotion recognition for real-life appli-cations is a challenging task. Human emotion expressions aresubtle, and can be conveyed by a combination of several emo-tions. In most existing emotion recognition studies, each audioutterance/video clip is labelled/classified in its entirety. However,utterance/clip-level labelling and classification can be too coarseto capture the subtle intra-utterance/clip temporal dynamics. Forexample, an utterance/video clip usually contains only a fewemotion-salient regions and many emotionless regions. In thisstudy, we propose to use attention mechanism in deep recurrentneural networks to detection the Regions-of-Interest (ROI) thatare more emotionally salient in human emotional speech/video,and further estimate the temporal emotion dynamics by aggre-gating those emotionally salient regions-of-interest. We comparethe ROI from audio and video and analyse them. We comparethe performance of the proposed attention networks with thestate-of-the-art LSTM models on multi-class classification task ofrecognizing six basic human emotions, and the proposed attentionmodels exhibit significantly better performance. Furthermore, theattention weight distribution can be used to interpret how anutterance can be expressed as a mixture of possible emotions.

cs.SD