AI Infrastructure

FlashAttention (FA1–FA4) in PyTorch - educational implementations focused on algorithmic differences [P]

I recently updated my FlashAttention-PyTorch repo so it now includes educational implementations of FA1, FA2, FA3, and FA4 in plain PyTor...

UMKC Announces New Master of Science in Artificial Intelligence

UMKC announces a new Master of Science in Artificial Intelligence program aimed at addressing workforce demand for AI expertise, set to l...

Siemens, NVIDIA hit chip verification milestone for AI

[2602.16340] The Implicit Bias of Adam and Muon on Smooth Homogeneous Neural Networks

This paper investigates the implicit bias of momentum-based optimizers like Adam and Muon in smooth homogeneous neural networks, extendin...

[2602.16336] HAWX: A Hardware-Aware FrameWork for Fast and Scalable ApproXimation of DNNs

HAWX introduces a hardware-aware framework for efficiently approximating deep neural networks (DNNs), achieving significant speedups whil...

[2602.16284] Fast KV Compaction via Attention Matching

The paper presents a novel approach for fast key-value (KV) compaction via Attention Matching, addressing the challenges of scaling langu...

[2602.16301] Multi-agent cooperation through in-context co-player inference

This paper explores multi-agent cooperation in reinforcement learning through in-context learning, demonstrating how sequence models can ...

[2602.16192] Revolutionizing Long-Term Memory in AI: New Horizons with High-Capacity and High-Speed Storage

This article discusses innovative approaches to long-term memory in AI, emphasizing the importance of retaining raw experiences for bette...

[2602.16179] EnterpriseGym Corecraft: Training Generalizable Agents on High-Fidelity RL Environments

The paper presents EnterpriseGym Corecraft, a novel high-fidelity reinforcement learning environment designed to train AI agents for gene...

[2602.16039] How Uncertain Is the Grade? A Benchmark of Uncertainty Metrics for LLM-Based Automatic Assessment

This article benchmarks various uncertainty metrics for LLM-based automatic assessment, highlighting the challenges of output uncertainty...

[2602.16197] ModalImmune: Immunity Driven Unlearning via Self Destructive Training

The paper presents ModalImmune, a training framework designed to enhance the resilience of multimodal systems against input channel loss ...

[2602.16155] Differentially Private Non-convex Distributionally Robust Optimization

This paper presents a novel approach to differentially private non-convex distributionally robust optimization (DRO), addressing challeng...

[2602.16092] Why Any-Order Autoregressive Models Need Two-Stream Attention: A Structural-Semantic Tradeoff

The paper explores the necessity of two-stream attention in any-order autoregressive models, highlighting a structural-semantic tradeoff ...

[2602.16052] MoE-Spec: Expert Budgeting for Efficient Speculative Decoding

The paper introduces MoE-Spec, a method for improving efficiency in speculative decoding of Large Language Models (LLMs) by optimizing ex...

[2602.16042] AI-CARE: Carbon-Aware Reporting Evaluation Metric for AI Models

The paper proposes AI-CARE, a carbon-aware evaluation metric for machine learning models, addressing the environmental impact of model tr...

[2602.16015] Geometry-Aware Uncertainty Quantification via Conformal Prediction on Manifolds

This paper introduces adaptive geodesic conformal prediction, a novel framework for uncertainty quantification on Riemannian manifolds, e...

[2602.15971] B-DENSE: Branching For Dense Ensemble Network Learning

The paper presents B-DENSE, a novel framework for improving dense ensemble network learning by leveraging multi-branch trajectory alignme...

[2602.15855] Kalman-Inspired Runtime Stability and Recovery in Hybrid Reasoning Systems

This article presents a framework for ensuring runtime stability and recovery in hybrid reasoning systems, emphasizing the importance of ...

Red and Blue States Alike Want To Limit AI in Insurance. Trump Wants To Limit the States.

A bipartisan movement is emerging across the U.S. to regulate AI in health insurance, challenging President Trump's push for less state o...

OpenAI taps Tata for 100MW AI data center capacity in India, eyes 1GW | TechCrunch

OpenAI partners with Tata Group to secure 100MW of AI data center capacity in India, aiming to expand to 1GW, enhancing enterprise AI ado...

OpenAI deepens India push with Pine Labs fintech partnership | TechCrunch

OpenAI partners with Pine Labs to enhance AI-driven payment solutions in India, aiming to streamline enterprise workflows and expand its ...

SDNY Addresses Privilege and Work Product Implications of Using Unsecured Public AI Tools

The SDNY ruled that AI-generated documents using unsecured public tools are not protected by attorney-client privilege, emphasizing the r...

AI weighs in on its own potential in fire and EMS

The article discusses how AI, particularly through insights from ChatGPT, is set to transform fire and EMS operations, focusing on govern...