Search: “inference”

WAND: Learning Robust Navigation under Complex Wind Disturbances and Dense Obstacles for Quadrotors

arXiv:2610.11809v1 Announce Type: new Abstract: Robust navigation in cluttered environments remains a fundamental challenge for quadrotors, particularly when strong wind disturbances arise, which pert

9 Oct 2026

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

What to Admit and How to Present: Governing Persistent Memory in LLM Agents

DGX agent

arXiv:2610.11188v1 Announce Type: new Abstract: Persistent memory can improve personalization in LLM agents but can also induce sycophancy and cross-domain leakage. We distinguish two governance decis

model-releasesarxiv-cs-ai

9 Oct 2026

When Citations Mislead? A Claim-Level Benchmark for Legal Hallucination Detection

model-releasesarxiv-cs-ai

arXiv:2610.10971v1 Announce Type: cross Abstract: Large language models are increasingly used in legal research and drafting, but they can still produce claims that sound convincing without being supp

9 Oct 2026

Where Draft Trees Lose Target Mass: Exit-Guided Speculative Decoding

arXiv:2610.11750v1 Announce Type: new Abstract: Tree-based speculative decoding verifies multiple draft continuations in one target-model pass, but finite trees built from draft scores face a fundamen

researcharxiv-cs-ai

9 Oct 2026

$2800 rig with 8x Radeon Pro V620 (256 GB VRAM) + custom vLLM fork = Qwen3.8-Flash-Next at 60 to 100 t/s decode and 3000+ t/s prefill

model-releasesr-localllama

Post title is slightly misleading, I don't think you can get these for $350 each anymore but they're still pretty cheap all things considered. They're Radeon Pro V620's which are older RDNA2 enterpris

A Bayesian Mirror Architecture for Emergent Consciousness: Circular Hierarchies, Self-Manifolds, and Hybrid Event-Self Binding

arXiv:2610.08792v1 Announce Type: new Abstract: We present a foundational formulation of the Bayesian Mirror Architecture (BMA), a self-referential generative framework in which sensory abstractions,

tutorialsarxiv-cs-lg

A Self-Pruning Transformer: Extreme KV-Cache Compression with Universal Attention

arXiv:2610.09051v1 Announce Type: new Abstract: The large KV-cache size of modern LLMs creates a barrier to efficient deployment. Recent work has explored replacing attention layers' RoPE positional e

A Vehicle-Integrated Approach to Digital Twin Deployment for Bridges Through Drive-By Sensing

arXiv:2610.08822v1 Announce Type: new Abstract: Ageing bridge infrastructure is a growing global concern, yet conventional Structural Health Monitoring (SHM) systems are costly and difficult to scale,

Adaptive Visual Token Reduction for Accelerated Image Understanding

arXiv:2610.09252v1 Announce Type: new Abstract: Large Vision-Language Models achieve strong VQA performance, but processing high-resolution, information-rich images requires substantial computation, m

Adversarial Images Hijack Web Agents from Visual Grounding to Browser Execution

arXiv:2610.09240v1 Announce Type: cross Abstract: Modern web agents built on large vision-language models process webpages, select relevant UI elements, and translate model outputs into browser action

Attention via Black-Box Vector Search

arXiv:2610.10135v1 Announce Type: cross Abstract: Sparse attention mechanisms estimate attention over n tokens using a small subset of keys. Many existing approaches use maximum inner product search (

[audio.cpp] Recent updates you might have missed: Higgs Audio TTS use 48% less VRAM (< 6GB), HTDemucs 2.2× faster, PocketTTS 2.2× faster on CPU, and WebUI generation history feature

Hi all, a bunch of performance improvements have been landed in audio.cpp. The biggest highlight is Higgs Audio TTS, which now runs with around 6 GB VRAM, a 48% reduction in peak memory usage compared

local-air-localllama

Bounded Autonomy and Verifiable Safety for Agentic AI Enabled Automation

model-releasesarxiv-cs-lg

arXiv:2610.08815v1 Announce Type: new Abstract: Agentic AI-enabled automation cannot be safely deployed in high-stakes environments on probabilistic reasoning alone. A recurring risk is epistemic drif

Bringing BNNs to Fast Event Processing

arXiv:2610.09873v1 Announce Type: new Abstract: Binary Neural Networks (BNNs) enable efficient deep learning deployment on resource constrained devices with weights and activations compressed to one b

chatgpt's new intelligent ui was reverse engineered in less than 24 hours, and apparently you can recreate it with local llms

model-releasesr-localllama

came across a pretty interesting technical breakdown of chatgpt's newly launched 'intelligent ui' feature, and thought this subreddit might find it interesting. for anyone unfamiliar with the concept,

CIRSeg: Coarse-to-Fine Intensity-Robust Liver Segmentation with Source-Free Continual Test-Time Adaptation

arXiv:2610.09784v1 Announce Type: new Abstract: Reliable liver segmentation in contrast-enhanced MRI is essential for quantitative hepatic assessment, treatment planning, and longitudinal disease moni

local-aiarxiv-cs-cv

CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models

arXiv:2610.08833v1 Announce Type: new Abstract: Masked diffusion language models (MDLMs) decode by repeatedly committing tokens to masked positions, but these commitments are usually irreversible. A t

researcharxiv-cs-cl

Collaborative Reasoning Distillation via Cross-Feedback and Coherent Curation

arXiv:2610.09587v1 Announce Type: cross Abstract: Reasoning capabilities are critical for advancing Large Language Models, yet current approaches either require massive computational budgets or strugg

researcharxiv-cs-cl

Contact-Aware Imitation Learning Through Contact Factorization

arXiv:2610.09533v1 Announce Type: new Abstract: Generalizable contact-rich manipulation requires robots to preserve intended task behavior while adapting its physical realization to changing contact c

Applications

CoreWeave makes the case for an open, full-stack AI cloud

The full-stack AI cloud market is about more than access to graphics processing units: it also encompasses the tools needed to improve models and agents in production. Providers are seeking to disting

applicationssiliconangle

Covariate-dependent Joint Modeling of Multivariate Ordinal Preferences and Its Connections with Comparison Models

arXiv:2610.09070v1 Announce Type: cross Abstract: Multivariate ordinal data along with covariates are commonly collected in problems ranging from alignment of language models with human preferences, a

safetyarxiv-cs-lg

DeltaSplat: Iterative Gaussian Refinement for Pose-Free Feed-Forward 3D Gaussian Splatting

arXiv:2610.09853v1 Announce Type: new Abstract: Pose-free feed-forward 3D Gaussian Splatting (3DGS) reconstructs a scene from sparse, unposed images in a single network pass, removing the need for cam

{Delta}WAM: Distilling Action Tangent Fields into World Action Models

arXiv:2610.09734v1 Announce Type: new Abstract: World Action Models (WAM) improve robot policies by augmenting sparse action supervision with dense future prediction. However, much of the predictable

local-aiarxiv-cs-ro

Detecting Adversarial Images through Response Profiles of Vision-Language Models

arXiv:2610.10436v1 Announce Type: new Abstract: Adversarial perturbations can alter the predictions of frozen vision-language models (VLMs) while leaving their confidence and image--text similarity pa

Do Better Visual Representations Always Lead to Better End-to-End Autonomous Driving?

arXiv:2610.09695v1 Announce Type: cross Abstract: Visual foundation models (VFMs) are increasingly integrated into end-to-end autonomous driving for their powerful representations, yet it remains uncl

safetyarxiv-cs-cv

EC-EarthFlow: Probabilistic emulation of daily transient global climate model simulations with flow matching

arXiv:2610.09715v1 Announce Type: new Abstract: We introduce EC-EarthFlow, a generative flow matching model that emulates simulations from the physical climate model EC-Earth3. The model is trained on

tutorialsarxiv-cs-lg

Edge Accuracy Is Not Enough: Why Dynamics-Learned Structure Fails to Transfer to Inverse Problems

arXiv:2610.10213v1 Announce Type: new Abstract: A natural strategy for inverse problems with scarce labelled data is to transfer relational structure learned from abundant forward-simulation data. We

Efficient 3D Gaussian Head Avatars for Edge Devices

model-releasesarxiv-cs-cv

arXiv:2610.09821v1 Announce Type: new Abstract: Generative 3D Gaussian head avatars provide high-quality, efficient rendering, but synthesising the Gaussian representation remains computationally expe

Efficient Provably Private Classification with a Tabular Foundation Model

arXiv:2610.10068v1 Announce Type: new Abstract: Tabular data underpin prediction and decision-making in medicine, finance, government and science, but often contain sensitive individual-level informat

tutorialsarxiv-cs-lg

Evaluating Trajectory Features for Routing Final-Layer Attention

model-releasesarxiv-cs-lg

arXiv:2610.09272v1 Announce Type: new Abstract: Attention routing requires a signal that predicts the value of attention on the current prefix. We evaluate whether hidden-state extrapolation error, cu

Event-Aligned Visual Action Reasoning for World Action Models

arXiv:2610.09427v1 Announce Type: new Abstract: World-Action Models (WAMs) utilize future visual prediction as an intermediate reasoning process to guide action generation. However, existing WAMs typi

tutorialsarxiv-cs-cv

Hardware

Expert Coupling in MoE Pretraining: Reducing All-to-All Overhead with Correlated Placement and Token Shuffling

arXiv:2610.09372v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) layers replace the feed-forward block of a Transformer with E expert networks, and each token is routed to k of these experts.

hardwarearxiv-cs-cl

FlashNeRD: Performance-First Contact-Rich Neural Robot Dynamics

arXiv:2610.09130v1 Announce Type: new Abstract: Compared with analytical physics, learned dynamics models promise robot simulation that is faster, inherently differentiable, and easily adaptable to re

FoldBack: Self-Correcting Masked Generative Policy for Long-Horizon Garment Folding

arXiv:2610.10462v1 Announce Type: new Abstract: We present FoldBack, a self-correcting masked generative policy for long-horizon garment folding. Existing long-trajectory policies may continue after a

For Those Who Believe in Faithfulness: Optimizing the Area Under Insertion and Deletion Curves for Ranking Relative Feature Importance

arXiv:2610.09844v1 Announce Type: cross Abstract: The adoption of machine learning for socially relevant tasks requires effective explainable artificial intelligence (XAI) methods to better understand

Applications

From Prompts to Trees: Effective LLM-Guided Tree Generation for Few-Shot Tabular Classification

arXiv:2610.10227v1 Announce Type: cross Abstract: While Large Language Models (LLMs) possess rich world knowledge and impressive generalization capabilities, their direct application to tabular data c

applicationsarxiv-cs-cl

Industry

Grok @Bot works well on your phone. Just download the app from Apple or Android. https://apps.apple.com/app/id6794501026

Grok @Bot works well on your phone. Just download the app from Apple or Android. https://apps.apple.com/app/id6794501026 This feels unreal Used Grok Bot today, mostly from my phone I got more done fro

industryelon-musk--x

Halogen + Qwen Flash Next keeps getting better

model-releasesr-localllama

With latest Halogen version update (0.17.2), decode is consistently at ~45 tps even at high context with Qwen 3.8 Flash Next on a 128GB Strix Halo. This is some great work u/peonist-ai. Have been pump

HAN-Mamba: Hierarchical Selective State Space Networks for Multi-Scale Financial Volatility Forecasting

model-releasesarxiv-cs-lg

arXiv:2610.10323v1 Announce Type: new Abstract: Short-horizon realized volatility forecasting requires the integration of market information that evolves at incompatible temporal resolutions, from sec

I built a local server that speaks the Ollama API - sessions survive crashes, and your chat doesn't wait behind your agents

Disclosure first: I'm one of the maintainers, this is my project. If you've pointed Open WebUI or a script at a local server, you know the failure modes: the server restarts and every conversation is

model-releasesr-ollama

Immiscible Diffusion Policy: Preserving Multimodal Robot Actions through Label-Free Noise Assignment

arXiv:2610.09369v1 Announce Type: cross Abstract: When diffusion policies were first introduced, they were expected to recover multi-modal action distributions. However, we find this expectation does

safetyarxiv-cs-lg

Inverting Multi-Vector Visual Document Indices

arXiv:2610.09920v1 Announce Type: cross Abstract: Prevailing multi-vector visual document retrievers store each page as about a thousand patch vectors, often in vector databases run by a third party.

Applications

KASALv2: Fully Automatic 3D Rotational Symmetry Classification and Axis Localization

arXiv:2610.09534v1 Announce Type: new Abstract: Rotational symmetry is an important prior in 6D pose estimation, improving pose accuracy and supporting symmetry-aware evaluation. However, current symm

applicationsarxiv-cs-cv

LiG-DETR: Local-in-Global Reassembly in Latent Space for Aerial Object Detection

model-releasesarxiv-cs-cv

arXiv:2610.09511v1 Announce Type: new Abstract: Aerial object detection faces substantial scale and density variations. Small objects are easily degraded by downsampling and feature compression, while

LRCC: Generalizing Low-Rank Compression with Conditional Computation

arXiv:2610.08858v1 Announce Type: new Abstract: Low-rank compression reduces the cost of pretrained language models by replacing linear transformations with low-rank factorizations. However, conventio

Agents

Markdown has become a universal representation of information between humans and agents. It preserves core primitives needed to express stru…

Markdown has become a universal representation of information between humans and agents. It preserves core primitives needed to express structure (headings, lists), allows agents to read the informati

agentsjerry-liu--x

Masked Feature Encoding for Large-Scale Whole Slide Image Representation

arXiv:2610.10225v1 Announce Type: new Abstract: Whole slide image (WSI) analysis in computational pathology follows a multiple instance learning (MIL) pipeline where patch embeddings are extracted ind

local-aiarxiv-cs-cv

MixedPEFT: Combining Multiple PEFT Methods with Mixed Objectives for Unsupervised Domain Adaptation

arXiv:2606.22272v2 Announce Type: replace Abstract: Applying pre-trained language models to new domains through full fine-tuning is computationally expensive and prone to catastrophic forgetting. To a