arXiv subjects

Manvinder Singh

Publications and source records attributed to Manvinder Singh.

3 recordsLinked to original sources

arXiv2025

Advancing Semantic Caching for LLMs with Domain-Specific Embeddings and Synthetic Data

This report investigates enhancing semantic caching effectiveness by employing specialized, fine-tuned embedding models. Semantic caching relies on embedding similarity rather than exact key matching, presenting unique challenges in balancing precision, query latency, and computational efficiency. We propose leveraging smaller, domain-specific embedding models, fine-tuned with targeted real-world and synthetically generated datasets. Our empirical evaluations demonstrate that compact embedding models fine-tuned for just one epoch on specialized datasets significantly surpass both state-of-the-art open-source and proprietary alternatives in precision and recall. Moreover, we introduce a novel synthetic data generation pipeline for the semantic cache that mitigates the challenge of limited domain-specific annotated data, further boosting embedding performance. Our approach effectively balances computational overhead and accuracy, establishing a viable and efficient strategy for practical semantic caching implementations.

cs.LG↗

arXiv2024

Patterns of Patterns II

Our earlier paper "Patterns of Patterns" combined three techniques from training, futures studies, and design in a design pattern called PLACARD that helps groups of people work together effectively. We used that pattern in five hands-on workshop case studies which took place at various locations in the US and the UK. This experience report documents what we learned, including the way our thinking about PLACARD evolved, together with additional patterns our work generated. We evaluate the reproducibility of our methods and results, and consider the broader economic implications of this way of working. We discuss implications of our prototyping work for the design of future platforms, drawing connections with recent developments in cognitive science and artificial intelligence. This positions our patterns of patterns as a toolkit for the design and governance of systems that combine social dynamics with technical components.

cs.SI↗

arXiv2020

An exact solution of the observable universe in Bianchi V space-time

In this paper we investigate an observable universe in Bianchi type V space-time by taking into account the cosmological constant as the source of energy. We have performed a $χ^{2}$ test to obtain the best fit values of the model parameters of the universe in the derived model. We have used two types of data sets, viz: i) 31 values of the Hubble parameter and ii) the 1048 Phanteon data set of various supernovae distance moduli and apparent magnitudes. From both the data sets, we have estimated the current values of the Hubble constant, density parameters $(Ω_{m})_{0}$ and $(Ω_Λ)_{0}$. The present value of deceleration parameter of the universe in derived model is obtained as $q_{0} = 0.59^{+0.04}_{-0.03}$ and $0.59^{+0.02}_{-0.03}$ in accordance with $H(z)$ and Pantheon data respectively. Also we observe that there is a signature flipping in the sign of deceleration parameter from positive to negative and transition red-shift exists. Thus, the universe in derived model represents a transitioning universe which is in accelerated phase of expansion at present epoch. We have estimated the current age of the universe $(t_{0})$ and present value of jerk parameter $(j_{0})$. Our obtained values of $t_{0}$ and $j_{0}$ are in good agreement with its values estimated by Plank collaborations and WMAP observations.

physics.gen-ph↗