Search: “inference”

Iterate on an LLM, Serve on an SLM

arXiv:2610.14576v1 Announce Type: new Abstract: As frontier LLMs gain more capabilities, the cost to run them also goes up. At the same time, small language models (SLMs) also gain more capabilities b

Hardware

Joint Subspace Sparse Attention for Efficient Long-Context Decoding

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

arXiv:2610.19026v1 Announce Type: new Abstract: Sparse attention reduces the KV-cache traffic that dominates long-context LLM decoding, but practical methods largely operate at coarse granularity. Tok

hardwarearxiv-cs-cl

15 Oct 2026

KoAGE: Do Large Language Models Know, Learn, and Coin Korean Neologisms?

arXiv:2610.19399v1 Announce Type: new Abstract: Korean neologisms arise through diverse mechanisms, a demanding test of language models beyond lexical memorization. To enable systematic evaluation, we

LaCache: Robust Semantic Caching for LLM Serving

arXiv:2608.01718v2 Announce Type: replace Abstract: Semantic caching, reusing responses to answer semantically similar queries, has seen growing adoption in LLM serving, offering faster responses and

researcharxiv-cs-ai

Large Decision Models: Repurposing Large Language Models as Single-Pass Decision Makers

arXiv:2610.15501v1 Announce Type: new Abstract: Large language models make decisions through parallel prefill and sequential decoding. Although effective, this approach incurs decoding costs and does

Lingdan-V2: Large language models for clinically grounded reasoning in traditional Chinese medicine

arXiv:2610.14705v1 Announce Type: cross Abstract: Large language models have advanced medical reasoning, but clinically grounded reasoning in traditional Chinese medicine (TCM) remains challenging bec

ListQA: A Benchmark for Evaluating List-Formatted Factual Knowledge Retrieval in Large Language Models

arXiv:2610.19165v1 Announce Type: cross Abstract: Many real-world questions demand not a single fact but an organized collection of them, yet existing factual knowledge benchmarks almost exclusively t

Making Language the Low-Rate Code of Vision: Successive Refinement from Words to Pixels

arXiv:2610.16552v1 Announce Type: new Abstract: An ordered visual code lets a short prefix stand in for an image, and the training loss decides what that prefix contains. Under pixel reconstruction th

researcharxiv-cs-ai

MemDuet: Learning Dual-Process Memory for Long-Context Reasoning

arXiv:2610.19928v1 Announce Type: new Abstract: Reasoning over infinitely long documents is becoming a reality with recurrent memory agents. However, increasing context length exacerbates the trade-of

safetyarxiv-cs-cl

Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs

arXiv:2610.19587v1 Announce Type: new Abstract: Social psychology has long described how people can be persuaded to interpret a harmful act as moral. Recent research suggests that Large Language Model

safetyarxiv-cs-cl

OntoAR-HAR: Ontology-Guided Agentic RAG with Personalized Memory for Human Activity Recognition

arXiv:2610.15292v1 Announce Type: new Abstract: LLMs increasingly support sensor-based human activity recognition (HAR) by linking motion signals with activity semantics and physical knowledge. Retrie

local-aiarxiv-cs-ai

PairIT: Autoregressive Transformers for Low-Data Molecule Optimization

arXiv:2610.19123v1 Announce Type: new Abstract: Generating novel molecules with desired properties is a central challenge in drug discovery. Although foundational work focused on de novo generation, p

local-aiarxiv-cs-lg

PerSo: Investigating Facets of LLMs' Social Intelligence in Real-World Persuasion Scenarios

arXiv:2610.20548v1 Announce Type: new Abstract: Understanding the social intelligence of large language models (LLMs) is increasingly important for applications involving human-AI interaction and mult

ProtModernBERT: Compute-Efficient Protein Language Modeling

model-releasesarxiv-cs-lg

arXiv:2610.19738v1 Announce Type: new Abstract: Large protein language models provide a powerful framework for learning sequence-function relationships and guiding experimental design. However, archit

Quantifying privacy risks in foundation model representations of brain MRI

model-releasesarxiv-cs-cv

arXiv:2610.20494v1 Announce Type: new Abstract: Pretrained representations are increasingly used to share and reuse information from medical images without exposing the original scans, but it is uncle

Refined Topology-Pass Scheduling: Damped Trust-Region Compensation for Dynamic k-NN Graphs in SE(3)-Equivariant Structure Prediction

arXiv:2610.19246v1 Announce Type: new Abstract: Equivariant structure-prediction pipelines alternate coordinate updates with the construction of a k-nearest-neighbour (kNN) graph whose membership is a

Remember, Revisit, Revise: Scaffold Memory for Long-Horizon 3D Geometry

arXiv:2610.19674v1 Announce Type: new Abstract: Geometry foundation models estimate camera poses and 3D scene structure from video sequences, but maintaining consistent reconstructions over long horiz

local-aiarxiv-cs-cv

Rethinking Prediction Horizons: Decoupling Model Queries from State Updates in Few-Step MeanFlow

arXiv:2610.20071v1 Announce Type: new Abstract: MeanFlow predicts interval-average velocities, but conventional few-step sampling ties the interval queried from the model to the duration of the state

REWIND: Uncertainty-Guided Trajectory Repair for Test-Time Reasoning

arXiv:2610.19561v1 Announce Type: new Abstract: An early mistake in the generated reasoning can constrain all subsequent reasoning. Repeated-generation methods, like MCMC-based approaches, attempt to

researcharxiv-cs-cl

SAGE: Pushing 2-Bit Vector-Quantized LLMs toward Real-World Usability

arXiv:2610.14745v1 Announce Type: new Abstract: Extreme low-bit weight-only quantization is an effective approach to reducing the deployment costs of large language models. Below 3 bits, scalar quanti

Scalable Robust MCTS with Anytime Deterministic Optimality Guarantees

arXiv:2610.15461v1 Announce Type: new Abstract: MDP planners often rely on nominal transition models that may differ from the true dynamics. Robust MDP planning accounts for this uncertainty through a

safetyarxiv-cs-ai

Scheduling Parallel Optical Circuit Switches for AI Training

arXiv:2603.07373v2 Announce Type: replace-cross Abstract: The rapid growth of AI training has dramatically increased datacenter traffic demand and energy consumption, which has motivated renewed inter

SchemaWeaver: Accelerating On-Device Mobile Agentic AI via Schema-Guided Parallel LLM Decoding

arXiv:2610.17436v1 Announce Type: new Abstract: Large language models (LLMs) on mobile devices can act as agents that call external software tools to complete complex tasks, but the major challenge of

local-aiarxiv-cs-ai

Applications

Schrodinger's Navigator: Imagining an Ensemble of Futures for Zero-Shot Object Navigation

arXiv:2512.21201v4 Announce Type: replace-cross Abstract: Zero-shot object navigation (ZSON) requires robots to find target objects in unseen environments without task-specific fine-tuning or pre-buil

applicationsarxiv-cs-cv

SWE-Spot: Building Small Repo-Experts with Repository-Centric Learning

arXiv:2601.21649v2 Announce Type: replace-cross Abstract: The deployment of coding agents in privacy-sensitive and resource-constrained environments drives the demand for capable open-weight Small Lan

Synthetic reaction data from transformation rules for graph-to-sequence pretraining in biodegradation pathway prediction

arXiv:2610.19640v1 Announce Type: new Abstract: The scarcity of observed biodegradation reactions constrains the training of product and pathway predictors. We present enviGraph, a framework that gene

Toward System One Clinical Intelligence: Parameter-Efficient Fine-Tuning and Counterfactual Explainability of Decision Foundation Models

arXiv:2610.15029v1 Announce Type: cross Abstract: Automated disease classification from patient-reported symptoms requires balancing the speed of lightweight discriminative models against the contextu

TreeSpark: Learning to Draft the Trees You Verify for Speculative Decoding

arXiv:2610.20175v1 Announce Type: new Abstract: Parallel drafters such as DFlash and DSpark reduce proposal latency, but early mismatches limit how much of each block can be accepted. Retaining altern

local-aiarxiv-cs-cl

Virtual Head Attention

arXiv:2610.17959v1 Announce Type: cross Abstract: The design of attention mechanisms in long-context large language models faces an ``impossible triangle'' among KV cache size, training and prefill FL

VisibilitySplat: Ray-Traced Occlusion for Relighting with 3D Gaussian Splatting

arXiv:2610.19847v1 Announce Type: new Abstract: We address inverse rendering with 3D Gaussian Splatting (3DGS) by introducing explicit visibility modeling for relighting as well as material and light

VITAL: From Visual Motion to Precise Action

arXiv:2610.20604v1 Announce Type: new Abstract: World-action models aim to turn visual anticipation into driving decisions, yet typically train action experts on trajectory representations defined ind

What Happens to the Goal? Distinct Failures of Policy Alignment in Goal Misgeneralization

arXiv:2610.19731v1 Announce Type: new Abstract: Goal misgeneralization is a common failure mode in which an agent behaves competently out of distribution but does not pursue the intended goal. However

safetyarxiv-cs-lg

When Skills Remain But Access Fails: Factorized Skill Memory for Continual Skill Internalization

arXiv:2610.15492v1 Announce Type: new Abstract: Skill internalization aims to convert externally specified procedural skills into persistent model capabilities, allowing language-model agents to execu

Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts

arXiv:2610.20067v1 Announce Type: new Abstract: Large Language Models (LLMs) frequently encounter conflicts between internal parametric knowledge and external contextual knowledge, yet the mechanisms

researcharxiv-cs-cl

Applications

1Step-WAM: Your World Action Model but Faster using 1-Step Action Generation

arXiv:2610.15625v1 Announce Type: new Abstract: World Action Models (WAMs) unify visual world modeling and action generation, yet their real-world deployment is heavily bottlenecked by iterative actio

applicationsarxiv-cs-cv

A Controlled Evaluation of Curvature Inspired Scalar Feature Scaling for Domain Generalization

arXiv:2610.18227v1 Announce Type: cross Abstract: Domain-specific curvature is an appealing idea for domain generalization (DG), but curvature learning can denote substantially different computations.

A Hardware-Partitioned micro-ROS Mini-Vehicle for Operator-Supervised Neural Driving

arXiv:2610.15966v1 Announce Type: new Abstract: We present a low-cost miniature Ackermann vehicle for embedded neural-driving research, organized as a hardware-partitioned perceive-decide-act agent: a

safetyarxiv-cs-ro

A Production-Grade, High-Efficiency Agentic RL Framework for Frontier LLMs with Validated Recipes

arXiv:2610.16352v1 Announce Type: new Abstract: Reinforcement learning (RL) has emerged as a critical phase in the post-training pipeline of large language models (LLMs). As the frontier of LLM capabi

safetyarxiv-cs-lg

An FPGA-Based Spiking Neural Network for Spatiotemporal Pattern Learning and Classification

arXiv:2610.17463v1 Announce Type: cross Abstract: Spiking Neural Networks (SNNs) are promising for energy-efficient spatiotemporal pattern recognition at the edge, but efficient deployment and on-devi

local-aiarxiv-cs-lg

Anchor-and-Harvest Curriculum: A Fully Unmasked Non-LLM Curriculum for Craftax

arXiv:2610.15328v1 Announce Type: new Abstract: We introduce Anchor-and-Harvest Curriculum (AHC), a domain-informed, rule-based training system for Craftax-Full. AHC combines achievement-oriented task

AnolisVLA: Lightweight Vision-Language-Action Model with Deep State Space Models for Mobile Manipulation

model-releasesarxiv-cs-ro

arXiv:2603.15046v2 Announce Type: replace Abstract: In this study, we address the problem of language-guided robotic manipulation, where a robot is required to manipulate a wide range of objects based

BAT3R: Robust Online 3D Reconstruction with Bayesian Adaptive State Updates

arXiv:2610.16330v1 Announce Type: new Abstract: Feed-forward transformer models have driven remarkable progress in 3D vision,yet their quadratic complexity renders them impractical for long video sequ

Applications

BrainAid: Efficient Brain Tumor Detection Using Lightweight CNNs and Explainable AI for Resource-Constrained Environments

arXiv:2610.15077v1 Announce Type: new Abstract: A brain tumor is an unusual growth of cells in the brain that can be cancerous or not, affecting normal brain function and causing serious risk to healt

applicationsarxiv-cs-cv

CacheLink: Functional KV Transport for Zero-Prefill Cross-Model Context Handoff

arXiv:2610.15509v1 Announce Type: new Abstract: KV caches make long-context serving efficient by reusing past computation, but model switching breaks this reuse and forces the receiver to prefill the

Compositional Average Velocity Modeling for Efficient Diffusion Offline Reinforcement Learning

arXiv:2610.14941v1 Announce Type: new Abstract: Diffusion-based policies have recently emerged as a powerful paradigm for offline reinforcement learning, but their reliance on multi-step denoising int

safetyarxiv-cs-lg

ConformalGuard: False-Positive-Controlled Action Gating for LM Classifiers

model-releasesarxiv-cs-lg

arXiv:2610.14365v1 Announce Type: new Abstract: Language model classifiers are increasingly deployed for high-stakes decisions such as phishing screening, content moderation, and fraud triage, yet val

Controllable Diverse Sampling through Pullback Directions

arXiv:2610.14294v1 Announce Type: cross Abstract: Conditional diffusion models can generate high-quality images that align closely with a given prompt. However, samples generated from different random

safetyarxiv-cs-cv

Correlation-Aware Reconstruction for Multivariate Time Series Anomaly Detection

model-releasesarxiv-cs-lg

arXiv:2610.15250v1 Announce Type: new Abstract: Reconstruction-based detectors for multivariate time series anomaly detection typically learn a single entangled representation of normal behavior and f