Abstract

Democratizing data access through natural language is a crucial goal for modern enterprises, but the practical adoption of Text-to-SQL is critically hindered by real-world complexities: 1. Obscure and large database schemas, 2. Ineffective retrieval of relevant tables and columns due to structured setting of schemas and vague user query, 3. Generation of syntactically or logically flawed SQL due to a lack of robust validation and correction mechanism. To address these systemic challenges, we introduce Reflect-SQL, a novel framework for Text to SQL, grounded in multi-stage self-reflection approach to develop understanding of obscure schema using a knowledge base, setup a process for effective retrieval and system to generate syntactically/semantically SQL. Instead of a single-pass attempt, our system employs an LLM-as-a-judge driven scoring mechanism within interconnected feedback loops to iteratively refine the results at every stage. A feedback-driven retrieval loop refines the user's natural language query, while a synthesis loop validates and corrects the SQL and finally, an entailment loop optimizes the end-to-end process and continuously enriches the knowledge base. By integrating these layers of reflection, Reflect-SQL bridges the critical gap between user intent and complex data. On the challenging BIRD benchmark, our framework achieves an execution accuracy of 72.03%, significantly outperforming state-of-the-art baselines, demonstrating a major leap in reliability for enterprise applications.

Abstract

LLMs have advanced code generation, but their use for generating microservices with explicit dependencies and API contracts remains understudied. We examine whether AI agents can generate functional microservices and how different forms of contextual information influence their performance. We assess 144 generated microservices across 3 agents, 4 projects, 2 prompting strategies, and 2 scenarios. Incremental generation operates within existing systems and is evaluated with unit tests. Clean state generation starts from requirements alone and is evaluated with integration tests. We analyze functional correctness, code quality, and efficiency. Minimal prompts outperformed detailed ones in incremental generation, with 50-76% unit test pass rates. Clean state generation produced higher integration test pass rates (81-98%), indicating strong API contract adherence. Generated code showed lower complexity than human baselines. Generation times varied widely across agents, averaging 6-16 minutes per service. AI agents can produce microservices with maintainable code, yet inconsistent correctness and reliance on human oversight show that fully autonomous microservice generation is not yet achievable.

Abstract

Benchmarks for large language models (LLMs) have progressed from snippet-level function generation to repository-level issue resolution, yet they overwhelmingly target implementation correctness. Software architecture tasks remain under-specified and difficult to compare across models, despite their central role in maintaining and evolving complex systems. We present ArchBench, the first unified platform for benchmarking LLM capabilities on software architecture tasks. ArchBench provides a command-line tool with a standardized pipeline for dataset download, inference with trajectory logging, and automated evaluation, alongside a public web interface with an interactive leaderboard. The platform is built around a plugin architecture where each task is a self-contained module, making it straightforward for the community to contribute new architectural tasks and evaluation results. We use the term LLMs broadly to encompass generative AI (GenAI) solutions for software engineering, including both standalone models and LLM-based coding agents equipped with tools. Both the CLI tool and the web platform are openly available to support reproducible research and community-driven growth of architectural benchmarking.

Abstract

Recent advances in agentic AI have shifted the focus from standalone Large Language Models (LLMs) to integrated systems that combine LLMs with tools, memory, and other agents to perform complex tasks. These multi-agent architectures enable coordinated reasoning, planning, and execution across diverse domains, allowing agents to collaboratively automate complex workflows. Despite these advances, evaluation and assessment of LLM agents and the multi-agent systems they constitute remain a fundamental challenge. Although various approaches have been proposed in the software engineering literature for evaluating conventional software components, existing methods for AI-based systems often overlook the non-deterministic nature of models. This non-determinism introduces behavioral uncertainty during execution, yet existing evaluations rely on binary task-completion metrics that fail to capture it. Evaluating agentic systems therefore requires examining additional dimensions, including the agent's ability to invoke tools, ingest and retrieve memory, collaborate with other agents, and interact effectively with its environment.

These challenges emerged during our ongoing industry collaboration with MontyCloud Inc., an industrial partner in the domain of Autonomous CloudOps, when we deployed an agentic system in production. These limitations surfaced during deployment, highlighting practical gaps in the current evaluation methods and the need for a systematic assessment of agent behavior beyond task outcomes. Informed by these observations and established definitions of agentic systems, we propose an end-to-end Agent Assessment Framework with four evaluation pillars encompassing LLMs, Memory, Tools, and Environment. We validate the framework on a representative Autonomous CloudOps use case, where experiments reveal behavioral deviations overlooked by conventional metrics, demonstrating its effectiveness in capturing runtime uncertainties.

Abstract

Lexical stress detection is vital for effective Computer-Assisted Language Learning (CALL) systems and is typically modeled as a binary task, labeling syllables as stressed or unstressed. However, detecting stress in non-native speech is challenging due to deviations from native (canonical) patterns and limited annotated data. We propose LexiRep, a one-shot framework utilizing linguistic knowledge for learning lexical stress representations. It combines a linguistically-guided contrastive framework-based discriminative stress representation learning with Improved Deep Embedded Clustering (IDEC) and iteratively refines the representations under linguistic constraints. We evaluated German and Italian non-native English speech with transfer learning from native data and observed that LexiRep achieves a 3% absolute improvement over the respective state-of-the-art while reducing model complexity by 99%. Further, LexiRep outperforms by 7.62% in an unsupervised setting, demonstrating the effectiveness of linguistically constrained contrastive learning for low-resource stress detection.

Abstract

Small centrifugal pumps, critical for rural and peri-urban water supply, often lack advanced monitoring. This study develops a low-cost operational Digital Twin (DT) for a 3.7 kW induction motor-driven pump at IIIT-Hyderabad (IIIT-H), optimised for resource-constrained settings with limited bandwidth and power. Using an ESP32 gateway, temperature, current, and vibration data are streamed to ThingSpeak, stored in MongoDB, and synchronised with a Simulink model via Eclipse Ditto. The system continuously compares simulated and actual pump data and raises alerts when abnormal behaviour is detected using a Root Mean Square Error (RMSE)-based anomaly detector. A dashboard was developed to enable remote motor operation and automated shutdown during unsafe conditions. The twin operates in a partially closed-loop mode: sensor streams continuously update the virtual model and generate alerts, while relay-based actuation enables automated start/stop motor operation. From August to November 2024, 160,000 minutely sampled records demonstrated an ≈8 -second end-to-end latency, which is characterised as near real-time and adequate for supervisory pump monitoring. This performance meets SCADA real-time standards, with a modelling error of approximately 5%. The system enabled anomaly detection, identifying faults hours in advance without labelled data, and achieved rapid operator adoption with minimal training. This open-source solution bridges a research gap for 3-5 HP pumps, offering a scalable blueprint for small utilities and supporting future enhancements like fully closed-loop control.

Abstract

Structural shifts pose a significant challenge for graph neural networks, as graph topology acts as a covariate that can vary across domains. Existing domain generalization methods rely on fixed structural augmentations or training on globally perturbed graphs, mechanisms that do not pinpoint which specific edges encode domain-invariant information. We argue that domain-invariant structural information is not rigidly tied to a single topology but resides in the consensus across multiple graph structures derived from topology and feature similarity. To capture this, we first propose EdgeMask-DG, a novel min-max algorithm where an edge masker learns to find worst-case continuous masks subject to a sparsity constraint, compelling a task GNN to perform effectively under these adversarial structural perturbations. Building upon this, we introduce EdgeMask-DG*, an extension that applies this adversarial masking principle to an enriched graph. This enriched graph combines the original topology with feature-derived edges, allowing the model to discover invariances even when the original topology is noisy or domain-specific. At equilibrium, the structural patterns that the task GNN relies upon are, by design, robust and generalizable. EdgeMask-DG* is the first to systematically combine adaptive adversarial topology search with feature-enriched graphs. We provide a formal justification for our approach from a robust optimization perspective. We demonstrate that EdgeMask-DG* achieves new state-of-the-art performance on diverse graph domain generalization benchmarks, including citation networks, social networks, and temporal graphs. Notably, on the Cora OOD benchmark, EdgeMask-DG* lifts the worst-case domain accuracy to {78.0%}, a {+3.8 pp} improvement over the prior state of the art (74.2%). The source code for our experiments can be found here: url{https://anonymous.4open.science/r/TMLR-EAEF/}

Abstract

Quantum information processing and computing tasks can be understood as quantum networks, comprising quantum states and channels and possible physical transformations on them. It is hence pertinent to estimate the change in informational content of quantum processes due to physical transformations they undergo. The physical transformations of quantum states are described by quantum channels, while the transformations of quantum channels are described by quantum superchannels. In this work, we determine fundamental limitations on how well the physical transformation on quantum channels can be undone or reversed, which are of crucial interest to design and benchmark quantum information and computation devices. In particular, we refine (strengthen) the quantum data processing inequality for quantum channels under the action of quantum superchannels. We identify a class of quantum superchannels, which appears to be the superchannel analog of subunital quantum channels, under the action of which the entropy of an arbitrary quantum channel is nondecreasing. We also provide a refined inequality for the entropy change of quantum channels under the action of an arbitrary quantum superchannel.

Abstract

The Log-Distance Path Loss (LDPL) model is a commonly used methodology to measure signal attenuation in Long Range Wide Area Networks (LoRaWAN), a wireless protocol primarily used for Internet of Things (IoT) applications. In this paper, we establish the limitations of this model even in near-ideal, line-of-sight (LoS) conditions, owing to significant deviations in empirical constants despite accounting for fading effects. The Path Loss Exponent (PLE) and the variance of shadow fading were calculated using Received Signal Strength Index (RSSI) measurements. They were obtained by sweeping the transmitter along radial and tangential directions with respect to a fixed receiver. At each transmitter position, multiple RSSI values were averaged to eliminate temporal fading. Both the transmitter and the receiver were secured at 0.5 m off the ground. The measurements were taken in an empty ground with transmitterreceiver LoS and minimal reflections, thus ensuring near-ideal experimental conditions. The PLE varied from 1.84 to 3.90 and the shadow fading varied from 0.5 dB to 5.8 dB, thus challenging the accuracy of RSSI applications, such as localization, that frequently use this model.

Abstract

FinFET technology has emerged as a solution to
mitigate the short-channel effects in CMOS technology. However, with continued technology scaling, leakage power is now emerging as a primary contributor to total power consumption. Additionally, aging effects, compounded by process, voltage, and temperature (PVT) variations, significantly degrade leakage behavior over time, resulting in high data variance and
posing challenges for conventional machine learning models to generalize aging-induced leakage trends. This paper proposes a physics-informed machine learning architecture based on a deaggregation-based methodology that effectively captures the complex, non-linear relationship between aging and leakage
power across PVT, incorporating both NBTI and HCI effects, by leveraging pattern-oriented knowledge. Experimental evaluations on 7nm and 10nm high-performance FinFET standard cells demonstrate mean absolute percentage error (MAPE) improvements of 52.71% and 54.44%, respectively, compared to stateof-the-art standalone boosting and deep learning models. The proposed method presents a robust and omputationally efficient alternative to traditional simulation-based approaches for agingaware power estimation in digital circuit design.