2890 papers
This paper proposes an LLM-based test case generation approach enhanced by a project memory repository that integrates historical data and release context to significantly improve risk coverage, reduce duplication, and increase defect discovery in large-scale software systems.
This paper establishes a preliminary MATLAB-based CNN baseline for binary malignant versus non-malignant breast ultrasound classification using the BrEaST dataset, achieving improved test accuracy while explicitly framing exploratory subtype labeling and explainability as future research directions rather than validated results.
This paper presents a modular differentiable programming framework that integrates interpretability directly into a continuous-time Markov chain model for clinical decision support by decomposing transition rates into clinically meaningful factors, computing their sensitivities via stochastic automatic differentiation, and generating traceable, plain-language explanations in real time.
This study compares the performance of NLTK, TextBlob, and spaCy in handling garden path sentences caused by functional shifts, finding that spaCy significantly outperforms the other two libraries in syntactic accuracy and offering insights for improving NLP tools, including those for African languages.
This paper presents a reference architecture and open-source implementation that bridges machine learning and operations research in supply chain planning by integrating conformal probabilistic forecasting with scenario-based stochastic routing, empirically demonstrating through a synthetic backtest that uncalibrated quantile forecasts lead to significant under-coverage and stockout risks, thereby validating the necessity of calibration-aware interfaces for reliable decision support.
This paper introduces a dematerialized teleoperation framework based on Kinematic Decoupling Control Theory that eliminates the need for physical kinesthetic twins by orthogonally decomposing human intent in information space, thereby achieving zero-shot generalization across diverse robot morphologies and overturning the long-held belief that physical force feedback is essential for intuitive robotic control.
Cethraian-X is an open-source, leakage-clean benchmark for multi-label chest X-ray classification using the NIH ChestX-ray14 dataset that establishes rigorous reproducibility standards through multi-seed evaluation, calibration analysis, and explainability tools to enable transparent comparison of future methods.
This paper presents a semantic-tracing framework for LLM-based multi-agent systems using LangChain, LangGraph, and LangSmith within the SHADOWAI-RISK prototype, demonstrating that integrating semantic evaluation significantly reduces semantic drift and constraint violations while incurring only a moderate latency increase and zero local cost.
This paper presents and evaluates an Automated Learning Engine (ALE) that advances beyond standard RAG by employing a multi-LLM orchestrated system with hybrid retrieval and state-machine workflows, demonstrating significantly improved factual accuracy, contextual relevance, and reduced hallucination rates in enterprise knowledge synthesis compared to baseline configurations.
Through empirical experiments on a fixed 24-million-token pool, this paper demonstrates that aggressive quality filtering under a fixed compute budget often harms model performance by forcing excessive data repetition, whereas retaining lower-quality data and oversampling high-quality subsets yields superior or equivalent results across various targets.