Author

Praveen Kumar Dhankar

View on Semantic Scholar

2 papers indexed here

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Preprint Aug 2026

Testing $f(Q)$ Gravity with DESI DR2 and Strong-Lensing Time Delays

Symmetric teleparallel gravity provides an alternative description of gravitation in which non-metricity replaces curvature and torsion. Its extension through $f(Q)$ gravity offers a different geometric description of the late-time expansion of the Universe and its accelerated phase. In this work, we investigate two $f(Q)$ models, a normalized power-law model and a square-root exponential model, and test their ability to describe the late-time expansion history. We constrain the model parameters through Markov chain Monte Carlo analyses using Cosmic Chronometer measurements, DESI DR2 baryon acoustic oscillations, strong-lensing time-delay observations, and three Type Ia supernova compilations, Pantheon$^+$, Union 3.0, and DES Y5. We compare both models with the flat $\Lambda$CDM model using the minimum $\chi^2$, Akaike information criterion, and Bayesian information criterion. The square-root exponential model provides a better statistical fit than $\Lambda$CDM for the combinations of Cosmic Chronometer, DESI DR2, and strong-lensing time-delay data with Pantheon$^+$ and Union 3.0, with improvements in both the goodness of fit and information criteria. The normalized power-law model remains statistically competitive with $\Lambda$CDM for the supernova-inclusive combinations, although the information criteria do not favor its additional parameter. We also determine the transition redshift from cosmic deceleration to acceleration for both models, obtaining consistent values across the different dataset combinations. The transition redshifts agree with observational estimates of the cosmic acceleration epoch. Overall, our results support $f(Q)$ gravity as a viable alternative to $\Lambda$CDM for explaining the late-time accelerated expansion of the Universe without requiring a cosmological constant.

Darshan Kumar, Saibal Ray, Fengge Zhang et al. · 0 citations

Conference Jul 2026

Parameter-Efficient Fine-Tuning of Transformer Models: An Empirical Study of LoRA Optimization for High-Cardinality Intent Classification

Although parameter-efficient fine-tuning significantly reduces the computing cost of deep models, default configurations are insufficient to perform as good as full fine-tuning for challenging large-cardinality intent detection problems with 77-151 intents. Thus, this work presents the S1 configuration that is proposed to remedy such performance degradation by defining the state-of-the-art low-rank adaptation. Instead of being constrained by the conventional formulation, this approach uses the minimal possible rank-8 adapter, full linear module coverage, and a learned learning rate. Extensive ablations offer two important discoveries that structural module coverage has more impact than mere adapter rank, and high learning rate is indispensable to provide enough convergence with the limited number of parameters. We show in the experiment that this configuration manages to restore the model performance to the 93.73% and 90.18% on Banking77 and CLINC150 respectively. S1 configuration attains the baseline parity, while only updating 1.56%-1.60% total trainable parameters with a maximal 2.96GB memory. This proves that it is possible to train a high-accuracy transformer on the hardware targeted at consumers, for example the NVIDIA RTX 5060 Ti.

Harsh Anand, Sofia Singh, Rahul Agrawal et al. · 2026 International Conferenc... · 0 citations