Linear Reservoir: A Diagonalization-Based Optimization
• Diagonalize recurrent matrix to transform reservoir dynamics into eigenbasis, eliminating costly matrix multiplication. • Achieves per-step computational complexity reduction fro
Machine-Learning
RA-QA: Towards Respiratory Audio-based Health Question Answering
• RA-QA dataset: 7.5M QA pairs, 60+ attributes, 3 question types. • Curated 11 respiratory audio datasets, first multimodal QA resource bridging audio and natural language. • Bench
Trojans in Artificial Intelligence (TrojAI) Final Report
• IARPA TrojAI program tackles AI trojan backdoor vulnerability in modern AI systems. • Found detection via weight analysis and trigger inversion techniques to identify hidden back
Simple Baselines are Competitive with Code Evolution
• Code evolution uses LLMs to mutate code, yet lacks baseline comparisons. • Authors test two simple baselines across math bounds, agentic scaffolds, and ML contests. • Baselines m
DistributedEstimator: Distributed Training of Quantum Neural Networks via Circuit Cutting
• Circuit cutting splits large quantum circuits into smaller subcircuits, enabling classical reconstruction of expectation values. • Prior studies focused on subcircuit counts and
Teaching machines to quiet the noise in quantum computers
• IIT Madras team uses machine learning to pinpoint noise sources in quantum computers. • Neural networks trained on synthetic data predict noise features with minimal accuracy los
Introducing Waypoint-1: Real-time interactive video diffusion from Overworld
• Waypoint‑1 is Overworld’s real‑time interactive video diffusion model, controllable via text, mouse, and keyboard. • Trained on 10,000 hours of diverse video‑game footage with co
Detecting faulty deployments: Our journey from unlabeled data to supervised learning
• Deployments drive 70% of incidents, making rapid fault detection critical for modern DevOps. • Datadog’s Automatic Faulty Deployment Detection uses APM telemetry to spot problema
Enhancing Neural Network Training at Yelp: Achieving 1,400x Speedup with WideAndDeep
• Reduced pCTR model training from 75 hours/epoch on 450M samples to under 1 hour on 2B samples. • Built ArrowStreamServer, an in‑house low‑latency Parquet streaming library, repla
Data Pipeline Challenges of Privacy-Preserving Federated Learning
• PPFL hides raw data from training org, preventing quality assessment and format validation. • Traditional preprocessing steps are often omitted in PPFL research, focusing solely
Machine Learning in Content Moderation at Etsy
• Etsy uses supervised ML to detect policy violations across 100M items. • Trust & Safety team combines community reports with automated removal. • Over 100,000 violations flagged
From Image Classification to Multitask Modeling: Building Etsy's Search by Image Feature
• Etsy launches image-based search on mobile, letting users find similar items via photos. • Feature uses machine‑learning embeddings to represent 100M+ listings in low‑dimensional
Modeling Sequences with Quantum States
• Introduces quantum-inspired methods for modeling sequential data via tensor networks. • Uses bipartite graph representations of matrices to visualize linear maps. • Applies quant
Understanding Entanglement With SVD
• SVD decomposes any matrix M into UDV†, revealing singular values and vectors. • Singular values, arranged in D, indicate the importance of each concept in M. • Left and right sin