Pix2Fact benchmark combining fine-grained visual grounding and web knowledge for evaluating vision-language models on high-resolution scenes.
Framework for explaining behavioral shifts in LLMs across scaling, fine-tuning, and RLHF interventions beyond static explainability methods.
Query-aware runtime memory routing system for LLM agents operating across extended context windows with cost-performance tradeoffs.
UG-Separation technique reducing compute costs in large recommendation models by decoupling user and group feature interactions.
Mathematical framework for optimal generative diffusion models using stochastic interpolants with efficient drift estimation.
Acoustic maps-based approach for detecting replay attacks in multi-channel speaker verification systems for voice assistants.
Context-aware explainability method using layer-wise integrated gradients to interpret transformer model predictions across network depth.
FT-Dojo benchmark environment enabling autonomous LLM fine-tuning through language agents, automating data curation and model configuration.
Research on timing side-channel vulnerabilities in LLM prefix caching optimization and mitigation strategies for multi-tenant systems.
Framework for routing inputs to experts with dynamic information selection, applicable to LLM systems with retrieved documents and tool outputs.
Medical video analysis pipeline using ResNet-50 for gastrointestinal pathology detection with class reweighting techniques.
Research on error bounds for physics-informed neural networks solving PDEs like Lyapunov and HJB equations with theoretical guarantees.
CRANE uses hidden Markov models to correct errors in raw nanopore sequencing electrical signals without basecalling.
Evaluates TabPFN foundation model on geotechnical site characterization tasks emphasizing uncertainty quantification and interpretability.
Praxium system diagnoses cloud microservice anomalies using AI-based telemetry analysis and dependency mapping for CI/CD environments.
Tutor-Student RL framework dynamically optimizes training curriculum for deepfake detection using agent-guided sample prioritization.
Analyzes why self-distillation degrades LLM reasoning in math tasks by suppressing epistemic verbalization of uncertainty.
Profiles energy consumption, performance, and privacy trade-offs for LLM deployment on edge devices with constrained memory and battery.
Demonstrates facial recognition systems exhibit demographic performance disparities, arguing aggregate accuracy inadequate for fairness evaluation in law enforcement.
Evaluates reliability and risk of AI systems in medication decision support, showing performance gaps between standard metrics and real-world deployment.
Proposes improved learning-to-defer systems that move beyond augmented-action surrogates for multi-expert decision routing.
Lightweight two-stage low-light image enhancement using algorithm-based preprocessing and depthwise-separable U-Net convolutions.
Physics-informed neural networks and Fourier neural operators for multi-scale wake modeling of floating offshore wind turbines.
JoyAI-Image unified multimodal foundation model combines spatially-enhanced MLLM with diffusion transformer for visual understanding and image generation.
Benchmarks JPEG decoder performance across multiple CPU architectures for ML data loaders, revealing limitations of single-thread microbenchmarks.
Comprehensive review of LLM applications in hardware design automation and electronic design, covering RTL code generation, testbenches, and security vulnerabilities.
Combines NeRF-based 3D scene representation with diffusion models for probabilistic 3D reconstruction addressing inherent uncertainty.
First online learning-to-defer algorithm for multiclass classification with dynamically varying expert pools and streaming data.
GAP method improves visual reasoning in multimodal LLMs by addressing feature-space mismatches in latent reasoning without external tools.
Theoretical analysis of computational sampling behavior in diffusion models using statistical mechanics, providing insights into convergence properties.
MeMo framework encodes new knowledge into dedicated memory models while keeping LLMs frozen, enabling efficient incorporation of domain-specific information without retraining.
Transformer-based physics simulator modeling diverse phenomena (cloth, fluids, solids) using unified Lagrangian particle representation and prediction-correction design.
Research on visuomotor policies for robotic manipulation using chunked action distributions and flow matching to improve trajectory coherence.
Charon: unified simulator for large-scale LLM training and inference with modular architecture for performance prediction.
SafeRun provides replay debugging and inline prevention for AI agents. SDK available in Python and TypeScript with sub-50ms API latency. Currently onboarding design partners.
Agent Chat Bridge enables async callbacks for AI IDE agents. Register timers, shell commands, or webhooks to resume agent sessions. Supports VS Code Copilot Chat and Windsurf Cascade.
Google AI Pro plan moves to credit-based usage system. Changes to free Gmail storage allocations and AI feature integration.
Opinion piece on Google competing with OpenAI in consumer AI. Discusses Google I/O conference and agentic AI era.
Ordo smart earbuds with built-in camera and on-device AI model. Voice commands, local processing, hands-free image/video capture.
TBN Protocol provides runtime governance infrastructure for AI agents. Includes automated guardrails and cost-control mechanisms with 14-step flow visualization.
Analysis of production issues in AI-built SaaS. Case study of tenant isolation bug missed in code review due to identical patterns.
SafeRun provides replay debugging and inline prevention for AI agents. SDK available in Python and TypeScript with sub-50ms API latency. Currently onboarding design partners.
Graph-based semantic memory system for AI coding agents. Local-first persistent memory using SQLite, llama.cpp, and BGE-M3. Compatible with Claude, Codex, ChatGPT.
Intuit laying off 3,000 employees (17% workforce) to refocus on AI integration in products. Corporate restructuring for AI investment.
AI system generates expert-level empirical software for scientific computing, addressing software bottlenecks in discovery cycles.
Nvidia reports record quarterly chip sales growth.
InferenceBench benchmark evaluates AI agents optimizing LLM serving workloads under compute constraints. Agents outperform baselines but underperform tuned hyperparameters.
Doubleword optimizes Mixture-of-Experts inference throughput by 15% via request reordering to reduce memory-bandwidth bottlenecks in batch processing.
Proposes using AI to find code quality patterns by analyzing swear word frequency in open source repositories. Experimental concept.
Stable Audio 3.0 open-weight music generation models trained on licensed data with variable-length generation up to 6 minutes, available on Hugging Face.