Indian Institute of Technology Delhi

Selected work

Representative Papers

Autonomous Microscopy Experiments through Large Language Model Agents

Dec 18, 2024arXiv.org

Existing self-driving laboratories (SDLs) rely on static experimental protocols, limiting their ability to emulate scientists’ adaptive reasoning and intuition in dynamic environments. Method: We propose AILA, the first large language model (LLM)-based autonomous agent system for end-to-end atomic force microscopy (AFM) experimentation—encompassing experimental design, execution, analysis, and closed-loop decision-making. Contribution/Results: We introduce AFMBench, the first benchmark for evaluating LLMs in AFM-driven scientific discovery, uncovering critical deficiencies in multi-agent coordination (73% failure rate), instruction following, and safety alignment, while empirically delineating LLMs’ scientific reasoning boundaries. Leveraging task-decomposition prompting, hardware interface integration, and a multi-agent architecture, AILA achieves autonomous AFM calibration, high-resolution feature identification, and nanomechanical property quantification. Results further reveal substantial accuracy degradation in foundational tasks (e.g., document retrieval), underscoring robustness and trustworthiness as central challenges in AI for Science.

2 citationsRead paper

Fair Distribution of Digital Payments: Balancing Transaction Flows for Regulatory Compliance

Nov 30, 2025arXiv.org

This study addresses the problem of fairly redistributing payment traffic under the regulatory constraint that no single UPI application may handle more than 30% of total transaction volume, modeling it as a Minimum Edge Activation Flow (MEAF) problem on a bipartite graph to minimize the number of additional payment applications users must install. The work formally defines this regulation-driven problem for the first time and proves its NP-completeness. To solve it efficiently, the authors propose a scalable two-stage decoupled allocation strategy (DTAS) that combines integer linear programming with heuristic methods, leveraging structural properties of traffic flows and capacity reuse mechanisms. Experimental results demonstrate that DTAS generates near-optimal, high-quality solutions within seconds on large-scale semi-synthetic networks, substantially improving regulatory compliance efficiency.

1 citations1 influentialRead paper

Graph Burning: Bounds and Hardness

Feb 29, 2024arXiv.org

This paper investigates the computational complexity and theoretical bounds of the graph burning number: given a graph $G$, one unburnt vertex is ignited per step, and its neighbors burn automatically in the next step; the goal is to minimize the number of steps required to burn the entire graph. Methodologically, the authors employ combinatorial graph theory, structural analysis of graph classes, and carefully constructed polynomial-time reductions. Their contributions include: (i) the first proof that graph burning remains NP-complete on connected cubic graphs and connected proper interval graphs; (ii) a tight additive-1 upper bound on the burning number for connected $P_k$-free graphs; and (iii) a systematic complexity classification of two natural variants—edge burning (igniting edges only) and total burning (igniting vertices or edges)—establishing their equivalence to the original vertex-burning problem. These results strengthen support for the conjecture that the burning number is at most $lceil sqrt{n} ceil$, and yield tight theoretical bounds for multiple fundamental graph classes while fully mapping the complexity landscape of the variants.

1 citations1 influentialRead paper

LaMSUM: Amplifying Voices Against Harassment through LLM Guided Extractive Summarization of User Incident Reports

Jun 22, 2024

To address the challenge of manually reviewing large-scale, code-mixed sexual harassment reports in India’s Safe City platform, this paper proposes the first LLM-driven extractive summarization framework tailored to this domain. Methodologically, it introduces a multi-model collaborative architecture integrating Llama, Mistral, and GPT-4o, enhanced by hierarchical text segmentation, prompt-engineered fine-grained extraction decisions, and an ensemble voting mechanism—effectively mitigating LLMs’ abstraction bias and context window limitations. Contributions include: (1) the first explainable and traceable extractive summarization system for code-mixed harassment reports; (2) state-of-the-art performance on the Safe City dataset, significantly outperforming existing baselines; and (3) generation of high-fidelity, structured event overviews that directly inform evidence-based policymaking and targeted anti-harassment interventions.

1 citationsRead paper

Use of Metric Learning for the Recognition of Handwritten Digits, and its Application to Increase the Outreach of Voice-based Communication Platforms

Jun 29, 2022The Compass

To address the low efficiency of paper-based data collection in resource-constrained settings, this work proposes a lightweight handwritten digit OCR method tailored for rural women’s Self-Help Groups (SHGs). We pioneer the adaptation of metric learning to low-quality, non-standard handwritten digit recognition, integrating CNN-based feature embedding with contrastive loss optimization to construct an end-to-end paper-to-digital pipeline augmented by Optical Mark Recognition (OMR). We publicly release the first large-scale, high-quality handwritten digit dataset specifically curated for this domain, along with corresponding lightweight models and open-source code. Evaluated on real-world rural survey forms, our method achieves 98.7% digit recognition accuracy. Deployed in a maternal and child health initiative across northern India, it has enabled over 3.9 million Interactive Voice Response (IVR) calls, substantially improving health information coverage, operational scalability, and data digitization efficiency.

1 citationsRead paper

Recent publications