Search: “inference”

MetaLearnNCA: Few-Shot Offline Meta-Learning via Interacting Neural Cellular Automata

arXiv:2610.08479v1 Announce Type: cross Abstract: Few-shot meta-learning traditionally formulates task adaptation either as analytical gradient descent through unrolled computational graphs or as metr

tutorialsarxiv-cs-ai

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Monte Carlo Estimation for KV Cache Eviction

DGX agent

arXiv:2610.07643v1 Announce Type: cross Abstract: Most KV-cache eviction methods ask, in effect, which memory appeared important while reading the prompt? We instead ask, which memory will matter whil

model-releasesarxiv-cs-ai

7 Oct 2026

More Value per Key: Asymmetric Sparse Attention for Faster LLM Decoding

arXiv:2610.04753v2 Announce Type: replace-cross Abstract: Autoregressive generation in Large Language Models (LLMs) is constrained by the memory and computational demands of attention mechanisms. Spar

researcharxiv-cs-ai

On the Intrinsic Limited Robustness of Latent-Based Watermarking

arXiv:2610.08178v1 Announce Type: new Abstract: Existing latent-based watermarking methods for diffusion models have overestimated their robustness to image distortions, including geometric transforma

Revisiting Numerical Forecasting Models for Language-Based Trajectory Prediction

arXiv:2610.07954v1 Announce Type: new Abstract: Language-based trajectory predictors represent coordinates as discrete tokens and learn auxiliary tasks such as destination and group reasoning. This fo

safetyarxiv-cs-cv

Safeguarding LLMs via Model-Agnostic Latent Safety Signals from Dark Knowledge

arXiv:2610.07532v1 Announce Type: cross Abstract: LLMs have advanced rapidly, raising growing concerns about their safety. Recent work has proposed approaches to detect and defend against attacks incl

SAFESHIELD: A Decision-Organization Framework for Deployment-Time Safety of Small Language Models

arXiv:2610.07276v1 Announce Type: cross Abstract: Deployment-time safety of language models is commonly implemented through runtime guardrails such as input moderation, routing, retrieval verification

SoloQ: Calibration-Free Quantization for Diffusion Language Models

arXiv:2610.07121v1 Announce Type: new Abstract: Diffusion large language models dLLMs) have emerged as a promising alternative to autoregressive language models through bidirectional diffusion-based t

StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models

arXiv:2610.07756v1 Announce Type: new Abstract: Vision-language-action (VLA) models increasingly rely on diffusion- or flow-matching-based action heads to generate continuous robot actions. These acti

local-aiarxiv-cs-ro

Structure-Aware Graph Abstention for Reliable Selective Forecasting

arXiv:2610.08322v1 Announce Type: new Abstract: Selective forecasting abstains on high-risk test windows under a retained-coverage budget. Existing gates such as TEM (Brusokas et al., 2025) score each

Applications

TacZero: Training-Free Peg Insertion Using a General-Purpose Vision-Language Model with Tactile Feedback

arXiv:2610.07621v1 Announce Type: new Abstract: Robots that autonomously determine their actions from language instructions and sensory observations could perform new contact-rich manipulation tasks w

applicationsarxiv-cs-ro

TAFFY: A Task-Adaptive Tabular Foundation Model with In-Context Diversity

arXiv:2610.07559v1 Announce Type: new Abstract: Recent progress in tabular foundation models suggests that training on synthetic tasks can substantially improve in-context learning capabilities, with

Test-Time Adaptation of Quantized ViTs via Single-Pass Quantizer-Aligned Recalibration

model-releasesarxiv-cs-ai

arXiv:2610.08358v1 Announce Type: cross Abstract: Post-training quantization is a standard route to fitting vision transformers (ViTs) into edge compute and memory budgets, yet quantized models become

Agents

Token-Efficient Multi-Agent Collaboration via System One-Guided Computational Division of Labor

arXiv:2610.08155v1 Announce Type: cross Abstract: Large language model (LLM)-based multi-agent systems (MAS) have become a promising paradigm for complex information-seeking and reasoning tasks by ena

agentsarxiv-cs-ai

Topology-Consistent Task Planning over Cellular Workflow Complexes for LLM-based Agents

arXiv:2610.07004v1 Announce Type: new Abstract: Task planning for LLM agents requires workflows that satisfy both user intent and complex sub-task dependencies. While existing planners work well for s

local-aiarxiv-cs-ai

Towards In-Parameter Memory Augmentation for Large Language Models

model-releasesarxiv-cs-cl

arXiv:2610.08630v1 Announce Type: new Abstract: Recently Large Language Models (LLMs) and LLM-based agents increasingly need to incorporate knowledge acquired after pretraining, e.g., domain facts, us

Towards One-for-All Foundation Model for Attributed Graph Clustering

arXiv:2610.07778v1 Announce Type: cross Abstract: Attributed graph clustering aims to discover node groups by jointly exploiting node attributes and graph topology, yet its unsupervised nature makes m

researcharxiv-cs-ai

Transferable Spatial Temporal Coherence Adversarial Attack on Black-Box Vision Language Models for Autonomous Driving

arXiv:2610.08331v1 Announce Type: cross Abstract: The rapid integration of Vision Language Models (VLMs) into sensitive systems introduces critical safety vulnerabilities that remain unexplored in exi

TRIAGE: Direction-Aware Mismatch Stabilization of Native NVFP4 Reinforcement Learning

arXiv:2610.07043v1 Announce Type: cross Abstract: Low-precision execution can substantially accelerate reinforcement learning (RL) for large language models, but discrepancies between learner and samp

TSRN-RTVD: Real-Time Video Deblurring System

arXiv:2610.08230v1 Announce Type: new Abstract: As video capture moves to handheld and edge devices, motion blur from camera shake has become a pervasive degradation that lowers perceptual quality and

local-aiarxiv-cs-cv

Two Vectors Replace In-Context Demos: Structured Task Adaptation via Embeddings

arXiv:2610.07572v1 Announce Type: new Abstract: In-context learning (ICL) adapts frozen large multimodal models (LMMs) to new tasks from a few demonstrations (demos), but re-encodes them at every quer

researcharxiv-cs-cl

Unbiased Reward Modeling from Implicit Feedback for LLM Alignment

model-releasesarxiv-cs-ai

arXiv:2603.23184v2 Announce Type: replace-cross Abstract: Despite the success of reinforcement learning from human feedback (RLHF), existing reward modeling methods largely rely on explicit feedback,

VeriFine: Scaling Verification for Self-Improvement in Embodied Reasoning

arXiv:2610.08761v1 Announce Type: new Abstract: Self-improving policies continually expose new failure patterns, changing what their judges must be able to verify. However, current fixed judges constr

What Frame-Level Labels Can and Cannot Do for Small-UAV Point Detection in Thermal Video

arXiv:2610.07705v1 Announce Type: cross Abstract: The growing use of unmanned aerial vehicles (UAVs) has increased the importance of image-based UAV detection. Learning-based detectors are trained on

local-aiarxiv-cs-ai

Why is ik_llama.cpp said to be faster than Mainline? On my hybrid multi-GPU rig, Mainline easily beats it

model-releasesr-localllama

I constantly see recommendations saying that ik_llama.cpp (ikawrakow's fork) is the undisputed king of hybrid CPU/GPU offloading and MoE performance. However, every time I benchmark it against mainlin

A Contrast-Source Inversion Scheme Based on Stochastic Optimization and Plug-and-Play Regularization

arXiv:2610.05130v1 Announce Type: cross Abstract: An electromagnetic inversion scheme that integrates stochastic optimization (STO) and plug-and-play (PNP) regularization into contrast-source inversio

A Fine-Grained Analysis of the LoRA Fine-Tuning Landscape with Implications for Data Selection

arXiv:2610.06542v1 Announce Type: new Abstract: Low-Rank Adaptation (LoRA) has become a standard approach for parameter-efficient fine-tuning, yet a fundamental practical question remains unresolved:

Agentic-ZTA: A Multi-Agent Architecture for Autonomous Zero Trust Enforcement

arXiv:2610.05782v1 Announce Type: cross Abstract: Agentic AI is emerging as a promising paradigm for automating complex cybersecurity decisions, yet its use in enforcing zero trust introduces signific

Industry

BattleBots builders turn to AI simulation to win fights before entering the box

AI infrastructure built on dense graphics processing unit clusters and high-speed AI networking is no longer reserved for frontier model labs. The same compute behind the industry’s push into physical

industrysiliconangle

Best practices for Amazon SageMaker HyperPod administration and governance

Learn how to administer Amazon SageMaker HyperPod through Amazon SageMaker Unified Studio while preserving cluster governance. This post shows platform teams how to design infrastructure boundaries, g

tutorialsaws-ml-blog

Conditional Trajectory Peaks: Single-Pass Multimodal Policies over Action Chunks

arXiv:2610.06104v1 Announce Type: new Abstract: Multimodal imitation learning requires diverse executable futures under the same observation and consistent behavior across replanning cycles. We presen

safetyarxiv-cs-ro

Coupling Noisy Pairwise Knowledge to the DAG Posterior for Causal Discovery

arXiv:2610.04559v1 Announce Type: new Abstract: External causal reports can improve structure learning from limited observations, but their reliability varies across sources and variable pairs. We int

Deep Prior Learning for Embodied Perception

model-releasesarxiv-cs-cv

arXiv:2610.05531v1 Announce Type: new Abstract: Embodied systems need geometric perception that exploits available observations beyond images alone. Recent feed-forward 3D models incorporate geometric

DiVeR: Decision-Critical Verifier Learning for VLA Test-Time Scaling

arXiv:2610.04933v1 Announce Type: cross Abstract: Scaling robot data and model capacity has improved Vision-Language-Action (VLA) policies, but further progress is constrained by the high cost of robo

tutorialsarxiv-cs-lg

Do More Modalities Always Help? A Geometric Perspective on Missing-Modality Robustness

model-releasesarxiv-cs-cv

arXiv:2610.04792v1 Announce Type: cross Abstract: Missing modality remains a longstanding challenge in multimodal learning. Existing methods typically address this issue through modality recovery or a

Does Explainability Survive Data Drift?

arXiv:2610.05379v1 Announce Type: new Abstract: Model performance monitoring is a standard practice in machine learning deployments. Detection performance is tracked continuously, and model decay is e

local-aiarxiv-cs-lg

DyQ-VLA: Temporal-Dynamic-Aware Quantization for Embodied Vision-Language-Action Models

arXiv:2603.07904v3 Announce Type: replace Abstract: Vision-language-action (VLA) models achieve favorable task performance, yet runtime errors in closed-loop execution evolve with alternating updates

EagleDepth: Efficient Fine-Grained Depth Estimation via Pixel Diffusion Decoder

arXiv:2610.04554v1 Announce Type: new Abstract: Recovering detailed geometry from high-resolution images is critical for precise perception of the surroundings and objects. However, existing methods w

researcharxiv-cs-cv

EDISCO: Equivariant DIScrete Diffusion for Euclidean Combinatorial Optimization

arXiv:2610.04953v1 Announce Type: new Abstract: Euclidean combinatorial optimization problems (ECOPs), such as the Traveling Salesman Problem (TSP) and Capacitated Vehicle Routing Problem (CVRP), poss

local-aiarxiv-cs-lg

Efficient Graph Generation via Direct Prediction and Flow Matching

arXiv:2610.05397v1 Announce Type: new Abstract: Generative modeling of graph-structured data is crucial for tasks ranging from drug discovery to social network simulation. Among these models, denoisin

tutorialsarxiv-cs-lg

Evolving in Thought Space: Training a Small Model at Test Time Unlocks Better Discoveries

arXiv:2610.06269v1 Announce Type: cross Abstract: Open-ended scientific discovery often requires repeatedly proposing and evaluating candidate solutions. LLM-based systems can support this process by

FLEX-WAM: Flexible Block-Causal World-Action Models for Long-Horizon Imagination and Planning

arXiv:2610.05483v1 Announce Type: new Abstract: World--action models (WAMs) promise a unified model that predicts action-conditioned futures, generates feasible actions, and supports planning in imagi

safetyarxiv-cs-ro

Gated Target Propagation for Compositional Generalization in Continual Learning

arXiv:2610.04649v1 Announce Type: new Abstract: Continual learning is typically framed as acquiring new knowledge without catastrophically forgetting previous tasks. However, a flexible continual lear

local-aiarxiv-cs-lg

GRAM: Correcting Frozen Time-Series Foundation Models via Graph-Retrieved Amplitude Memory

arXiv:2610.04827v1 Announce Type: new Abstract: Time-series foundation models (TSFMs) enable zero-shot forecasting through large-scale cross-domain pretraining, while retrieval augmentation further im

Applications

Graph Data Augmentation via Contrastive Generator Inversion (exttt{DCBA})

arXiv:2610.05653v1 Announce Type: new Abstract: Graphs provide a natural representation of many complex systems, ranging from social platforms to ecosystems. However, the development of graph-based ma

applicationsarxiv-cs-lg

Green-Routed Neural Operators: Physics Determines Where the Network Reads

arXiv:2610.05337v1 Announce Type: new Abstract: We identify a mismatch between the physical role of transport fields in many PDEs and their usual role in neural operators: PDEs use them to select read

H-JEPA: End-to-End Learning of Hierarchical World Models for Visual Planning

arXiv:2610.06805v1 Announce Type: new Abstract: Long-horizon planning with latent world models requires reasoning across timescales and levels of abstraction. Existing task-agnostic JEPA world models

Harnessing Multimodal Large Language Models for Training-Free Human-Object Interaction Detection

model-releasesarxiv-cs-cv

arXiv:2610.06394v1 Announce Type: new Abstract: Human-object interaction (HOI) detection aims to localize human-object pairs and recognize their interactions. Traditional supervised methods perform st