Hardware/Chips arXiv cs.AI

Chimera: Neuro-Symbolic Attention Primitives for Trustworthy Dataplane Intelligence

Proposes Chimera, a framework that maps neuro-symbolic attention primitives onto programmable dataplanes for line-rate, low-latency traffic analysis.

AI/ML arXiv cs.AI

NeuroSymActive: Differentiable Neural-Symbolic Reasoning with Active Exploration for Knowledge Graph Question Answering

Introduces NeuroSymActive, a framework for Knowledge Graph Question Answering that combines differentiable neural-symbolic reasoning with active exploration.

AI/ML arXiv cs.AI

AdvSynGNN: Structure-Adaptive Graph Neural Nets via Adversarial Synthesis and Self-Corrective Propagation

Presents AdvSynGNN, a graph neural network architecture designed for resilient node-level representation learning in the presence of structural noise.

AI/ML arXiv cs.AI

SubQuad: Near-Quadratic-Free Structure Inference with Distribution-Balanced Objectives in Adaptive Receptor framework

Introduces SubQuad, a scalable pipeline for adaptive immune repertoire analysis using near-subquadratic retrieval and GPU-accelerated kernels.

AI/ML arXiv cs.AI

Pre-Deployment Complexity Estimation for Federated Perception Systems

Proposes a pre-deployment framework to estimate learning complexity in federated perception systems for edge AI deployment.

Cybersecurity arXiv cs.AI

DocShield: Towards AI Document Safety via Evidence-Grounded Agentic Reasoning

Introduces DocShield, a framework for detecting text-centric image forgeries using visual-logical co-reasoning and the RealText-V1 dataset.

Hardware/Chips Hacker News

The Visual 6502

A look at the Visual 6502 simulator, which provides a visual representation of the 6502 processor's internal state.

Other Hacker News

A24 Is Copyright Striking Backrooms Artwork Older Than the Movie

Discussion regarding A24 issuing copyright strikes against Backrooms artwork created before the movie's release.

Tech Business/VC TechCrunch

Meet the judges who will crown Australia’s next breakout startup

TechCrunch announces its Startup Battlefield competition coming to Australia in partnership with Stripe.

AI/ML arXiv cs.AI

Matching Ranks Over Probability Yields Truly Deep Safety Alignment

Researchers introduce PRESTO, a method to strengthen deep safety alignment in LLMs against Rank-Assisted Prefilling (RAP) attacks.

AI/ML arXiv cs.AI

Dominant vs. Dominated: Concept-Level Generative Collapse in Diffusion Models

The paper identifies 'Dominant-vs-Dominated' (DvD) imbalance in diffusion models and introduces DominanceBench to study multi-concept generation failure modes.

AI/ML arXiv cs.AI

ArenaRL: Scaling RL for Open-Ended Agents via Tournament-based Relative Ranking

ArenaRL is proposed as a reinforcement learning paradigm using tournament-based relative ranking to scale agents for open-ended tasks.

AI/ML arXiv cs.AI

Generative Semantic Multi-Object Tracking: A Large-Scale Benchmark and an MLLM-Driven Reasoning Framework

The researchers introduce Grand-SMOT, a large-scale benchmark, and LLMTrack, an MLLM-driven framework for generative semantic multi-object tracking.

AI/ML arXiv cs.AI

Learning About Learning: A Path from Spin Glasses to Artificial Intelligence

A pedagogical approach to teaching the Hopfield model, linking statistical mechanics and AI for undergraduate physics curricula.

AI/ML arXiv cs.AI

Geometric Attention: A Regime-Explicit Operator Semantics for Transformer Attention

The paper introduces Geometric Attention (GA), a formal operator semantics for Transformer attention to enable principled comparison of attention mechanisms.

AI/ML arXiv cs.AI

Comparative evaluation of training strategies using partially labelled datasets for segmentation of white matter hyperintensities and stroke lesions in FLAIR MRI

A study evaluating training strategies using partially labelled datasets for the segmentation of brain lesions in FLAIR MRI images.

AI/ML VentureBeat

Microsoft launches new in-house AI models it says cut costs up to 89% versus OpenAI

Microsoft released new in-house AI models (MAI-Image-2.5-Pro and MAI-Voice-2-Flash) that significantly reduce GPU costs and improve efficiency over OpenAI's frontier models for specific enterprise tasks.

AI/ML arXiv cs.AI

Fara-1.5: Scalable Learning Environments for Computer Use Agents

Researchers introduced FaraGen1.5, a scalable data pipeline for computer use agents, and released Fara1.5, a family of native computer use agents based on Qwen3.5 under an MIT license.

AI/ML arXiv cs.AI

Distributed Optimization via Energy Conservation Laws in Dilated Coordinates

A new double-loop distributed optimization method is proposed to achieve an O(k^-2) aggregate-objective rate, overcoming discretization barriers in primal-dual flows.

AI/ML arXiv cs.AI

Leveraging ChatGPT's Multimodal Vision Capabilities to Rank Satellite Images by Poverty Level: Advancing Tools for Social Science Research

This research demonstrates that multimodal LLMs like ChatGPT can rank satellite images by poverty level with accuracy comparable to domain experts, aiding social science research.