Sagar Sen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sagar Sen has a total of 115 co-authors. Here are the top co-authors by number of shared publications:
- Arda Goknil
SINTEF Digital
13 shared publications - Phu Hong Nguyen
SINTEF
12 shared publications - Erik Johannes Husom
SINTEF Digital
11 shared publications Benoit Baudry
Inria, Rennes, France
10 shared publications- Simeon Tverdal
SINTEF Digital
8 shared publications Jean-Marie Mottu
AtlanModels Team (Inria, IMT-A, LS2N), University of Nantes, Nantes, France
7 shared publications
Latest Publications
- Open Access
- 25 October 2026
- Published by Springer Nature
- p. 100-110
From Smell to Solution: Multi-agent LLMs Analyzing Cyclic Dependencies for LLM-Enabled Refactoring
Cyclic dependencies are a persistent architectural smell that undermines modularity, blurs ownership, and elevates maintenance risk. While Large Language Models (LLMs) provide refactoring abilities, existing tools struggle to address smells without intervention evidence. We introduce a multi-agent LLM framework for cyclic dependency analysis, which transforms static dependency evidence into structured, verifiable explanations of architectural entanglement. The approach constructs a structural dependency graph, detects cyclic regions via strongly connected components (SCC), extracts cycles as minimal witnesses of cyclicity, then produces grounded analyses. These are produced by a hierarchical team of LLM agents that describe the position, structure, and architectural impact of each cycle, clarifying why it exists and where it can be safely resolved. Evaluation on seven open-source Python projects shows that evidence-linked analyses can guide LLM-driven edits, reducing strongly connected components size and improving structural quality.
- 25 October 2026
- Published by Springer Nature
- p. 100-110
Open AccessFrom Smell to Solution: Multi-agent LLMs Analyzing Cyclic Dependencies for LLM-Enabled Refactoring
Cyclic dependencies are a persistent architectural smell that undermines modularity, blurs ownership, and elevates maintenance risk. While Large Language Models (LLMs) provide refactoring abilities, existing tools struggle to address smells without intervention evidence. We introduce a multi-agent LLM framework for cyclic dependency analysis, which transforms static dependency evidence into structured, verifiable explanations of architectural entanglement. The approach constructs a structural dependency graph, detects cyclic regions via strongly connected components (SCC), extracts cycles as minimal witnesses of cyclicity, then produces grounded analyses. These are produced by a hierarchical team of LLM agents that describe the position, structure, and architectural impact of each cycle, clarifying why it exists and where it can be safely resolved. Evaluation on seven open-source Python projects shows that evidence-linked analyses can guide LLM-driven edits, reducing strongly connected components size and improving structural quality.
- 28 August 2026
- Published by Springer Nature
- p. 72-86
Multi-agent Generative AI on Digital Twins: A Learning Feedback Loop Framework for Monitoring and Explainability
- 28 August 2026
- Published by Springer Nature
- p. 72-86
Multi-agent Generative AI on Digital Twins: A Learning Feedback Loop Framework for Monitoring and Explainability
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 23-33
Distilling Ensemble Intelligence into Explainable Anomaly Detection Models
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 23-33
Open AccessDistilling Ensemble Intelligence into Explainable Anomaly Detection Models
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 161-165
Code Wars: Adversarial Self-Play for Evolving Software Validation Tools
Self-play has reshaped the landscape of AI, producing agents like AlphaGo and AlphaZero that outmatch human mastery through recursive learning. In this work, we ask: can machines play against themselves not to win games, but to strengthen software? We explore this idea in the realm of software validation, where large language models (LLMs) trained on vast oceans of code hold within them entire worlds of logic, error, and repair. Our proof-of-concept system features three LLM-driven agents: an attacker that crafts flawed code, a defender that hunts for vulnerabilities, and a judge that weighs their wits. In this evolving dance of sabotage and defense, the agents sharpen each other, creating a self-sustaining arms race of refinement. In our preliminary evaluation of adversarial self-play for technical debt detection, we observe improvement in accuracy from 19.73% to 49.66% on a benchmark dataset. Our findings offer an early indication that ideas inspired by self play may contribute to the development of software validation tools.
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 161-165
Open AccessCode Wars: Adversarial Self-Play for Evolving Software Validation Tools
Self-play has reshaped the landscape of AI, producing agents like AlphaGo and AlphaZero that outmatch human mastery through recursive learning. In this work, we ask: can machines play against themselves not to win games, but to strengthen software? We explore this idea in the realm of software validation, where large language models (LLMs) trained on vast oceans of code hold within them entire worlds of logic, error, and repair. Our proof-of-concept system features three LLM-driven agents: an attacker that crafts flawed code, a defender that hunts for vulnerabilities, and a judge that weighs their wits. In this evolving dance of sabotage and defense, the agents sharpen each other, creating a self-sustaining arms race of refinement. In our preliminary evaluation of adversarial self-play for technical debt detection, we observe improvement in accuracy from 19.73% to 49.66% on a benchmark dataset. Our findings offer an early indication that ideas inspired by self play may contribute to the development of software validation tools.
- Open Access
- 19 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet Technology
- Vol. 26 (2) , 1-34
Unsupervised Learning and Process Analysis for Sensor Data Validation in the IIoT
Integrating Artificial Intelligence (AI) with the Industrial Internet of Things (IIoT) has transformed industrial processes, enhancing productivity, quality control, and operational efficiency. However, ensuring the precision and reliability of sensor-generated data remains a critical challenge due to the evolving nature of industrial processes and the limitations of conventional validation methods. Traditional rule-based and supervised learning approaches struggle to adapt to process shifts, drifts, and novel anomalies, making sensor data validation an ongoing issue. This article introduces UDAVA (Unsupervised Learning Approach using Process Mining for Sensor Data Validation in IIoT), a novel AI-driven pipeline designed to automate the identification of reference patterns in sensor data and validate subsequent production cycles by recognizing deviations from expected behaviors. UDAVA employs a multi-stage process that includes preprocessing sensor data, clustering recurring patterns, and assessing deviations. It supports semi-supervised learning by integrating manual labels where available, improving interpretability and accuracy. One of UDAVA’s key strengths lies in its ability to extract features from sensor data rather than relying on raw time series similarity, making it robust against noise and diverse process variations. Additionally, UDAVA integrates process mining techniques—process discovery and conformance checking—to enhance its ability to detect even subtle anomalies and deviations in industrial workflows. We conduct a comprehensive evaluation of UDAVA using three industrial datasets, demonstrating its effectiveness in identifying high-level process behaviors, detecting process shifts and drifts, and ensuring data validation across multiple production cycles. The results highlight UDAVA ’s adaptability across different industrial processes, making it a valuable tool for optimizing operations and ensuring sensor data reliability in IIoT environments.
- 19 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet Technology
- Vol. 26 (2) , 1-34
Open AccessUnsupervised Learning and Process Analysis for Sensor Data Validation in the IIoT
Integrating Artificial Intelligence (AI) with the Industrial Internet of Things (IIoT) has transformed industrial processes, enhancing productivity, quality control, and operational efficiency. However, ensuring the precision and reliability of sensor-generated data remains a critical challenge due to the evolving nature of industrial processes and the limitations of conventional validation methods. Traditional rule-based and supervised learning approaches struggle to adapt to process shifts, drifts, and novel anomalies, making sensor data validation an ongoing issue. This article introduces UDAVA (Unsupervised Learning Approach using Process Mining for Sensor Data Validation in IIoT), a novel AI-driven pipeline designed to automate the identification of reference patterns in sensor data and validate subsequent production cycles by recognizing deviations from expected behaviors. UDAVA employs a multi-stage process that includes preprocessing sensor data, clustering recurring patterns, and assessing deviations. It supports semi-supervised learning by integrating manual labels where available, improving interpretability and accuracy. One of UDAVA’s key strengths lies in its ability to extract features from sensor data rather than relying on raw time series similarity, making it robust against noise and diverse process variations. Additionally, UDAVA integrates process mining techniques—process discovery and conformance checking—to enhance its ability to detect even subtle anomalies and deviations in industrial workflows. We conduct a comprehensive evaluation of UDAVA using three industrial datasets, demonstrating its effectiveness in identifying high-level process behaviors, detecting process shifts and drifts, and ensuring data validation across multiple production cycles. The results highlight UDAVA ’s adaptability across different industrial processes, making it a valuable tool for optimizing operations and ensuring sensor data reliability in IIoT environments.
- 20 November 2025
- Published by Springer Nature
- p. 334-352
Detecting Technical Debt in Source Code Changes Using Large Language Models
- 20 November 2025
- Published by Springer Nature
- p. 334-352
Detecting Technical Debt in Source Code Changes Using Large Language Models
- Top Cited PapersOpen Access
- 18 November 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet of Things
- Vol. 6 (4) , 1-35
Sustainable LLM Inference for Edge AI: Evaluating Quantized LLMs for Energy Efficiency, Output Accuracy, and Inference Latency
Deploying Large Language Models (LLMs) on edge devices presents significant challenges due to computational constraints, memory limitations, inference speed, and energy consumption. Model quantization has emerged as a key technique to enable efficient LLM inference by reducing model size and computational overhead. In this study, we conduct a comprehensive analysis of 28 quantized LLMs from the Ollama library, which applies by default Post-Training Quantization (PTQ) and weight-only quantization techniques, deployed on an edge device (Raspberry Pi 4 with 4 GB RAM). We evaluate energy efficiency, inference performance, and output accuracy across multiple quantization levels and task types. Models are benchmarked on five standardized datasets (CommonsenseQA, BIG-Bench Hard, TruthfulQA, GSM8K, and HumanEval), and we employ a high-resolution, hardware-based energy measurement tool to capture real-world power consumption. Our findings reveal the trade-offs between energy efficiency, inference speed, and accuracy in different quantization settings, highlighting configurations that optimize LLM deployment for resource-constrained environments. By integrating hardware-level energy profiling with LLM benchmarking, this study provides actionable insights for sustainable AI, bridging a critical gap in existing research on energy-aware LLM deployment.
- 18 November 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet of Things
- Vol. 6 (4) , 1-35
Top Cited PapersOpen AccessSustainable LLM Inference for Edge AI: Evaluating Quantized LLMs for Energy Efficiency, Output Accuracy, and Inference Latency
Deploying Large Language Models (LLMs) on edge devices presents significant challenges due to computational constraints, memory limitations, inference speed, and energy consumption. Model quantization has emerged as a key technique to enable efficient LLM inference by reducing model size and computational overhead. In this study, we conduct a comprehensive analysis of 28 quantized LLMs from the Ollama library, which applies by default Post-Training Quantization (PTQ) and weight-only quantization techniques, deployed on an edge device (Raspberry Pi 4 with 4 GB RAM). We evaluate energy efficiency, inference performance, and output accuracy across multiple quantization levels and task types. Models are benchmarked on five standardized datasets (CommonsenseQA, BIG-Bench Hard, TruthfulQA, GSM8K, and HumanEval), and we employ a high-resolution, hardware-based energy measurement tool to capture real-world power consumption. Our findings reveal the trade-offs between energy efficiency, inference speed, and accuracy in different quantization settings, highlighting configurations that optimize LLM deployment for resource-constrained environments. By integrating hardware-level energy profiling with LLM benchmarking, this study provides actionable insights for sustainable AI, bridging a critical gap in existing research on energy-aware LLM deployment.
- Open Access
- 18 March 2025
- Published by Association for Computing Machinery (ACM) in Journal of Data and Information Quality
- Vol. 17 (1) , 1-28
3D-DaVa: Enhancing 3D Point Cloud Data Reliability for Industrial Applications
The escalating incorporation of three-dimensional (3D) point cloud data across industrial applications highlights the necessity of assuring its reliability. The error-prone process of object digitization, the large data volumes, and equipment inaccuracies can lead to the degradation of 3D point cloud data quality by introducing varying degrees of noise, outliers, and missing values. Therefore, there is a pressing need for a generally applicable, comprehensive, and robust solution to effectively validate the integrity of point cloud data and assess its quality. Such a solution would empower professionals to rely on this data for critical decision-making, covering applications within many domains, such as manufacturing, automotive, and robotics. In this article, we propose, apply, and assess the 3D Point Cloud Data Validation (3D-DaVa) pipeline, an automated 3D data validation system incorporating statistical and machine learning techniques. The pipeline takes a point cloud and its reference as input and outputs accuracy, validity, and completeness scores. We demonstrated the efficacy of 3D-DaVa through a rigorous evaluation using both real-world manufacturing data and openly available data, where we deliberately introduced distortions covering five distortion levels by simulating common inaccuracies. The data quality assessment results obtained by varying distortion levels reveal a decreasing trend with a distortion increase, thus underscoring the 3D-DaVa’s capability to quantify such deviations accurately.
- 18 March 2025
- Published by Association for Computing Machinery (ACM) in Journal of Data and Information Quality
- Vol. 17 (1) , 1-28
Open Access3D-DaVa: Enhancing 3D Point Cloud Data Reliability for Industrial Applications
The escalating incorporation of three-dimensional (3D) point cloud data across industrial applications highlights the necessity of assuring its reliability. The error-prone process of object digitization, the large data volumes, and equipment inaccuracies can lead to the degradation of 3D point cloud data quality by introducing varying degrees of noise, outliers, and missing values. Therefore, there is a pressing need for a generally applicable, comprehensive, and robust solution to effectively validate the integrity of point cloud data and assess its quality. Such a solution would empower professionals to rely on this data for critical decision-making, covering applications within many domains, such as manufacturing, automotive, and robotics. In this article, we propose, apply, and assess the 3D Point Cloud Data Validation (3D-DaVa) pipeline, an automated 3D data validation system incorporating statistical and machine learning techniques. The pipeline takes a point cloud and its reference as input and outputs accuracy, validity, and completeness scores. We demonstrated the efficacy of 3D-DaVa through a rigorous evaluation using both real-world manufacturing data and openly available data, where we deliberately introduced distortions covering five distortion levels by simulating common inaccuracies. The data quality assessment results obtained by varying distortion levels reveal a decreasing trend with a distortion increase, thus underscoring the 3D-DaVa’s capability to quantify such deviations accurately.
- Open Access
- 19 December 2024
- Published by Association for Computing Machinery (ACM) in Journal of Data and Information Quality
- Vol. 16 (4) , 1-3
Editorial: Special Issue on Software Engineering and AI for Data Quality
This editorial summarizes the content of the Special Issue on Software Engineering and AI for Data Quality of the Journal of Data and Information Quality (JDIQ).
- 19 December 2024
- Published by Association for Computing Machinery (ACM) in Journal of Data and Information Quality
- Vol. 16 (4) , 1-3
Open AccessEditorial: Special Issue on Software Engineering and AI for Data Quality
This editorial summarizes the content of the Special Issue on Software Engineering and AI for Data Quality of the Journal of Data and Information Quality (JDIQ).
- Open Access
- 19 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 28-37
Data Quality as a Service Framework for AI-Enabled Industrial Internet of Things
- 19 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 28-37
Open AccessData Quality as a Service Framework for AI-Enabled Industrial Internet of Things