arXiv subjects

Ishan Singh

Publications and source records attributed to Ishan Singh.

4 recordsLinked to original sources

arXiv2026

Study of discrete boundary layer suction for transition delay

In the present study we have assessed the effect of suction through discrete perforations on the transition of the Blasius boundary layer. In particular, the intermittency and the spectra of the streamwise velocity fluctuations, obtained using hot-wire anemometry, were used to evaluate the optimum suction, i.e., one that provides the greatest delay in transition. The displacement thickness Reynolds number at suction was varied between 1400 and 1600, while the number of rows of holes was either four or ten. The suction holes had diameter of the order of the displacement thickness of the boundary layer, a practical size from manufacturing and maintenance perspectives. As the suction strength was increased the transition front moved downstream, with the maximum delay in transition occurring at the optimum suction. When the suction was increased further, the transition front moved upstream again. The results, supported by a scaling analysis, showed that suction volume flow rate is a better indicator of the effects of suction on the transition delay than suction velocity. In particular, suction through different numbers of rows of perforations produces largely similar effects on the boundary layer transition, provided the suction volume flow rate remains the same. However, suction through fewer rows provided a slightly greater delay in transition in addition to a slightly lower optimum suction volume flow rate. The results also showed that suction applied at a lower Reynolds number provides a substantially greater transition delay but at slightly higher values of the normalized optimum suction volume flow rate.

physics.flu-dyn↗

arXiv2026

An Empirical Study of Automating Agent Evaluation

Agent evaluation requires assessing complex multi-step behaviors involving tool use and intermediate reasoning, making it costly and expertise-intensive. A natural question arises: can frontier coding assistants reliably automate this evaluation process? Our study shows that simply prompting coding assistants is insufficient for this task. Without domain-specific evaluation knowledge, frontier coding assistants achieve only a 30% execution success rate and produce over-engineered evaluations averaging 12+ metrics per agent, indicating that strong coding ability does not automatically translate to reliable agent evaluation. We introduce EvalAgent, an AI assistant that automates the end-to-end agent evaluation pipeline. EvalAgent encodes evaluation domain expertise as evaluation skills (procedural instructions, reusable code and templates, and dynamically retrieved API documentation) that compose into a trace-based pipeline producing complete evaluation artifacts including metrics, executable code, and reports. To systematically assess generated evaluations, we introduce a meta-evaluation framework alongside AgentEvalBench, a benchmark comprising 20 agents, each paired with evaluation requirements and test scenarios. We further propose the Eval@1 metric to measure whether generated evaluation code both executes and yields meaningful results on the first run. Our experiments show that EvalAgent produces focused evaluations, improving Eval@1 from 17.5% to 65%, and achieving 79.5% human expert preference over baseline approaches. Further ablation studies show that evaluation skills are critical for handling complex evaluation: removing them causes Eval@1 to drop significantly from 65% to 30%.

cs.CL↗

arXiv2025

Keyword search is all you need: Achieving RAG-Level Performance without vector databases using agentic tool use

While Retrieval-Augmented Generation (RAG) has proven effective for generating accurate, context-based responses based on existing knowledge bases, it presents several challenges including retrieval quality dependencies, integration complexity and cost. Recent advances in agentic-RAG and tool-augmented LLM architectures have introduced alternative approaches to information retrieval and processing. We question how much additional value vector databases and semantic search bring to RAG over simple, agentic keyword search in documents for question-answering. In this study, we conducted a systematic comparison between RAG-based systems and tool-augmented LLM agents, specifically evaluating their retrieval mechanisms and response quality when the agent only has access to basic keyword search tools. Our empirical analysis demonstrates that tool-based keyword search implementations within an agentic framework can attain over $90\%$ of the performance metrics compared to traditional RAG systems without using a standing vector database. Our approach is simple to implement, cost effective, and is particularly useful in scenarios requiring frequent updates to knowledge bases.

cs.IR↗

arXiv2022

A Numerical Study of Lid Driven Cavity with Mixed Convection

Direct Numerical Simulation have been carried out for a two dimensional flow in a Lid driven cavity at Reynolds number 5000 and Prandtl number 7 with water as the working fluid. Both the side walls of the enclosure are insulated(i.e. adiabatic boundary condition), while the bottom plate is at higher temperature and the top wall is at colder temperature. Effects of heating of the bottom wall and movement of the top lid have been investigated by conducting numerical simulations at different Richardson numbers by varying from low and moderate magnitudes within the limits of Boussinesq-approximation. Three standard cases has been compared, in the first case heating effects are not taken into account and only the flow due to shear action of the plate is studied. In the second case only the heating effects are taken into account and shear effects are neglected. In the third case effects of both heating and shear action is taken into consideration(i.e. mixed convection). Drag force on the moving plate is calculated in all the three cases and effect of temperature on the drag force is studied. For running the above simulation a code has been developed which is validated by comparing the results with Ghia et al for non-heating case.

physics.flu-dyn↗