local ai

AIGaitor: Privacy-preserving and cloud-free motion analysis for everyone, using edge computing

arXiv:2605.21421v1 Announce Type: new Abstract: Motion capture is the gold standard for measuring human movement, but clinical use remains limited by cost, technical complexity, and privacy concerns.

b9255

Release b9255 of llama.cpp features a Hexagon HMX quantized matmul rework (#23368), including updates to debug logging, dequantization logic using HVX vectors, removal of non-pipelined quantization op

local-aillama-cpp-releases

21 May 2026

b9257

llama.cpp release b9257 optimizes operations in the Vulkan IM2COL shader and improves code formatting . The release includes prebuilt binaries for multiple platforms including macOS, Linux, Android, a

b9259

Release b9259 of llama.cpp fixes a nullptr crash in the speculative common module's get_devices_str function (PR #23386). The fix addresses an issue where ggml_backend_dev_by_name appends a nullptr se

b9260

Release b9260 of llama.cpp includes OpenCL backend refactoring that improves initialization, GPU identification, and performance by caching global memory size in device context. llama.cpp enables LLM

b9263

Release b9263 of llama.cpp includes a merge of HunyuanOCR into HunyuanVL with fixes to OCR vision precision. The update consolidates OCR functionality into the HunyuanVL projector while maintaining co

b9264

b9264 is a llama.cpp release that includes improvements to HunyuanVL model support, merging HunyuanOCR functionality and fixing vision precision issues. This build represents an intermediate developme

b9266

b9266 is a release tag for llama.cpp, a project for LLM inference in C/C++. This release represents a specific build version of the llama.cpp framework, which enables LLM inference with minimal setup

b9270

Release b9270 adds support for the HybridDNATokenizer used by the Carbon-3B model family, implementing a new BPE pre-type for tokenizing DNA sequences. The tokenizer handles DNA k-mers with fixed 6-me

b9273

Release b9273 of llama.cpp introduces support for the HybridDNATokenizer with new pre-type and dispatched tokenization logic, alongside pure helper functions for DNA k-mer processing and conversion ut

b9275

Release b9275 of llama.cpp includes optimization of the Metal concat kernel and fixes to the GGML_OP_SET kernel threads . The release extends test coverage for copy operations with different source an

Closed-form predictive coding via hierarchical Gaussian filters

arXiv:2605.20293v1 Announce Type: new Abstract: Predictive coding (PC) offers a local and biologically grounded alternative to backpropagation in the training of artificial neural networks, yet to dat

Conformal Selective Acting: Anytime-Valid Risk Control for RLVR-Trained LLMs

arXiv:2605.20270v1 Announce Type: new Abstract: A local specialist LLM, fine-tuned with reinforcement learning from verifiable rewards (RLVR) on operator-local data, is installed in a regulated organi

Control FLUX.2 with reference images instead of training a LoRA — demo

This post demonstrates using reference images with control maps in FLUX.2 instead of training LoRAs, allowing users to fuse a reference image with pose or edge control data while preserving identity a

Creé mi propio agente de IA local centrado en Ollama y la ejecución de tareas prácticas.

This post describes a user's experience building a local AI agent using Ollama, an open-source tool for running large language models locally. The article likely covers the practical implementation st

local-air-ollama

Diffusion Models Memorize in Training -- and Generalize in Inference

arXiv:2603.13419v2 Announce Type: replace Abstract: Diffusion models generalize well in practice. However, an optimal diffusion model fully memorizes the training data and therefore fails to generaliz

DNACHUNKER: Learnable Tokenization for DNA Language Models

arXiv:2601.03019v4 Announce Type: replace-cross Abstract: DNA language models are increasingly used to represent genomic sequence, yet their effectiveness depends critically on how raw nucleotides are

local-aiarxiv-cs-cl

E-ReCON: An Energy- and Resource-Efficient Precision-Configurable Sparse nvCIM Macro for Conventional and Spiking Neural Edge Inference

arXiv:2605.20717v1 Announce Type: cross Abstract: This work presents E-ReCON, a 16 Kb energy and resource-efficient digital compute-in-memory (DCIM) macro based on a compact 3T1R ReRAM bitcell for edg

End-to-End Unmixing with Material Prompts for Hyperspectral Object Tracking

arXiv:2605.20569v1 Announce Type: new Abstract: Hyperspectral imagery encodes rich material properties that can improve tracking robustness under appearance ambiguity, illumination change, and backgro

Explainable AI: Context-Aware Layer-Wise Integrated Gradients for Explaining Transformer Models

arXiv:2602.16608v2 Announce Type: replace Abstract: Transformer models achieve state-of-the-art performance across domains and tasks, yet their deeply layered representations make their predictions di

local-aiarxiv-cs-cl

Federated Learning of Nonlinear Temporal Dynamics with Graph Attention-based Cross-Client Interpretability

arXiv:2602.13485v2 Announce Type: replace Abstract: Networks of modern industrial systems are increasingly monitored by distributed sensors, where each system comprises multiple subsystems generating

FLUME-FNO: data-efficient and scalable prediction of 3D wind and temperature fields in unseen urban morphologies

arXiv:2503.19708v2 Announce Type: replace-cross Abstract: Urban microclimate, encompassing wind and temperature fields shaped by building geometry, significantly impacts energy consumption, pedestrian

FTerViT: Fully Ternary Vision Transformer

arXiv:2605.21171v1 Announce Type: new Abstract: Ternary Vision Transformers offer substantial model compression, however state-of-the-art methods only ternarize the encoder layers, leaving patch embed

HDMoE: A Hierarchical Decoupling-Fusion Mixture-of-Experts Framework for Multimodal Cancer Survival Prediction

arXiv:2605.20891v1 Announce Type: new Abstract: Multimodal survival prediction, a crucial yet challenging task, demands the integration of multimodal medical data (eg Whole Slide Images (WSIs) and Gen

I added a visual Fold feature for organizing large ComfyUI workflows

A developer added a visual fold feature to ComfyUI, a node-based visual programming environment used for Stable Diffusion workflows. This feature helps organize complex workflows with many nodes by al

I'm excited about the new @amd Ryzen AI Halo because we need more local hardware for AI builders! There's something fun and exciting about b…

I'm excited about the new @amd Ryzen AI Halo because we need more local hardware for AI builders! There's something fun and exciting about building on your own machines rather than sending to the clou

local-aiclem-delangue--x

IndusAgent: Reinforcing Open-Vocabulary Industrial Anomaly Detection with Agentic Tools

arXiv:2605.20682v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) have shown remarkable capability in bridging visual perception and textual reasoning, enabling zero-shot unders

Intent-First Aerial V2V for Tactical Coordination and Separation: Protocol and Performance Under Density and Disturbance

arXiv:2605.20595v1 Announce Type: new Abstract: Dense low-altitude aerial operations require more than pre-flight route coordination and last-resort collision avoidance. Once aircraft are airborne, di

local-aiarxiv-cs-ro

Interaction Locality in Hierarchical Recursive Reasoning

arXiv:2605.20784v1 Announce Type: cross Abstract: Spatial reasoning requires both location-bound computation and location-invariant structure: agents must make local moves while preserving route, obje

Lip sync and Lora

LoRA models can be used with image-to-video workflows to create automatic lip-syncing effects by controlling camera movements and instructing the model to synchronize audio with mouth movements. This

LoCar: Localization-Aware Evaluation of In-Vehicle Assistants through Fine-Grained Sociolinguistic Control

arXiv:2605.21086v1 Announce Type: new Abstract: While Large Language Models (LLMs) are increasingly integrated into in-vehicle conversational systems, identifying the optimal model remains challenging

local-aiarxiv-cs-cl

LOSCAR-SGD: Local SGD with Communication-Computation Overlap and Delay-Corrected Sparse Model Averaging

arXiv:2605.20866v1 Announce Type: new Abstract: Communication is a major bottleneck in distributed learning, especially in large-scale settings and in federated learning environments with slow links.

LTX 2.3 + LTX Director is a Huge improvement

LTX-2.3 is a multimodal video generation model developed by Lightricks that generates synchronized audio and video in a single forward pass at resolutions up to 4K at 50 frames per second. The model i

Mechanisms of Misgeneralization in Physical Sequence Modeling

arXiv:2605.20299v1 Announce Type: new Abstract: Generative sequence models are often trained to plan motion in physical domains, from robotics to mechanical simulations. When constructing a dataset to

Modality-Decoupled Online Recursive Editing

arXiv:2605.20273v1 Announce Type: new Abstract: Online model editing for multimodal large language models (MLLMs) requires assimilating a stream of corrections under tight compute and memory budgets.

Multi-needle Localization for Pelvic Seed Implant Brachytherapy based on Tip-handle Detection and Matching

arXiv:2509.17931v2 Announce Type: replace Abstract: Accurate multi-needle localization in intraoperative CT images is crucial for optimizing seed placement in pelvic seed implant brachytherapy. Howeve

OmniISR: A Unified Framework for Centralized and Federated Learning via Intermediate Supervision and Regularization

arXiv:2605.20276v1 Announce Type: new Abstract: The global deployment of edge intelligence operates across heterogeneous legal frameworks. While some regions permit centralized learning (CL) via cloud

On prem.

On prem. I'm excited about the new @amd Ryzen AI Halo because we need more local hardware for AI builders! There's something fun and exciting about building on your own machines rather than sending to

local-aiclem-delangue--x

One-Step Distillation of Discrete Diffusion Image Generators via Fixed-Point Iteration

arXiv:2605.21484v1 Announce Type: new Abstract: Discrete diffusion models excel at visual synthesis but rely on slow, iterative decoding. Existing single-step distillation methods attempt to bypass th

Online 3D Multi-Camera Perception through Robust 2D Tracking and Depth-based Late Aggregation

arXiv:2509.09946v2 Announce Type: replace Abstract: Multi-Target Multi-Camera Tracking (MTMC) is an essential computer vision task for automating large-scale surveillance. With camera calibration and

Optimized Federated Knowledge Distillation with Distributed Neural Architecture Search

arXiv:2605.21322v1 Announce Type: new Abstract: Federated Learning (FL) enables collaborative model training without centralizing data. However, real-world deployments must simultaneously address stat

OSGNet with MLLM Reranking @ Ego4D Episodic Memory Challenge 2026

arXiv:2605.20818v1 Announce Type: new Abstract: In this report, we present our champion solutions for the Natural Language Queries and GoalStep tracks of the Ego4D Episodic Memory Challenge at CVPR 20

PaintCopilot: Modeling Painting as Autonomous Artistic Continuation

arXiv:2605.20941v1 Announce Type: new Abstract: We present PaintCopilot, a co-creative neural painting assistant that models painting as an open-ended autoregressive artistic behavior conditioned on e

Quant.npu: Enabling Efficient Mobile NPU Inference for on-device LLMs via Fully Static Quantization

arXiv:2605.20295v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed on mobile devices, where Neural Processing Units (NPUs) necessitate fully static quantization for

R2AoP: Reliable and Robust Angle of Progression Estimation from Intrapartum Ultrasound

arXiv:2605.21099v1 Announce Type: new Abstract: Accurate estimation of the Angle of Progression (AoP) from intrapartum transperineal ultrasound is critical for objective assessment of labor progressio

RePCM: Region-Specific and Phenotype-Adaptive Bi-Ventricular Cardiac Motion Synthesis

arXiv:2605.21237v1 Announce Type: new Abstract: Cardiac motion over a cardiac cycle is crucial for quantifying regional function and is strongly affected by cardiovascular diseases. Since temporally d

Scale-Calibrated Median-of-Means for Robust Distributed Principal Component Analysis

arXiv:2605.20681v1 Announce Type: cross Abstract: Distributed principal component analysis (PCA) produces node-level estimates of both a mean vector and a principal subspace. Robustly aggregating thes

Semantic Granularity Navigation in Image Editing

arXiv:2605.21190v1 Announce Type: new Abstract: Despite the generative capabilities of diffusion and flow models, real-image editing remains constrained by a persistent trade-off between semantic edit