Search: “inference”

Self-Reconstruction Dynamics for Autoencoder Reconstruction Refinement

arXiv:2609.32268v1 Announce Type: new Abstract: Standard autoencoder (AE) inference uses a single encoder-decoder pass, though the latent may not be optimal for each sample under a fixed decoder. We a

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Hardware

SparSP: Exploiting Communication Sparsity for Sequence-Parallel Video DiTs

DGX agent

arXiv:2609.32197v1 Announce Type: cross Abstract: Diffusion Transformers have become the dominant architecture for video generation. Their substantial computational cost motivates scaling inference ac

hardwarearxiv-cs-lg

29 Sept 2026

Hardware

Systematic Exploration of Multi-core Architectures for Efficient LLM Serving using WaferAI-SIM

arXiv:2510.05632v2 Announce Type: replace-cross Abstract: With the widespread adoption of Large Language Models (LLMs), the demand for high-performance LLM inference services continues to grow. Multi-

hardwarearxiv-cs-lg

Towards Transparent Diagnostics: Investigating Architectural Trade-offs and Explainability in Malaria Detection

arXiv:2609.31682v1 Announce Type: new Abstract: More than 80 countries have reported malaria cases with 610 thousand deaths and are projected to increase. Identifying malaria early and accurately help

researcharxiv-cs-cv

Two-Sample Testing for Inhomogeneous Random Graphs in Non-Integral L_r Norms

arXiv:2609.33968v1 Announce Type: cross Abstract: Testing whether two populations of networks share the same edge probabilities is a basic problem in network inference. How hard it is depends on the n

Hardware

UnStep: Training-Free Acceleration of Causal Video Diffusion with Fewer Steps Than Distillation

arXiv:2609.32518v1 Announce Type: new Abstract: Distilling bidirectional multi-step video diffusion transformers into few-step causal models has become a common approach for streaming video generation

hardwarearxiv-cs-cv

When Keywords Drop but Classifiers Hold: Soft Refusals under KV Cache Compression

arXiv:2609.31678v1 Announce Type: new Abstract: KV cache compression is widely used for long context LLM inference under memory constraints, while deployed systems typically score refusals after gener

Applications

When Less Compute Is More: Adaptive Early Exit Improves Pretrained Outlier Detection

arXiv:2609.32898v1 Announce Type: new Abstract: Pretrained tabular foundation models process every dataset at a fixed depth, with inference costs growing with dataset size. To address this, we present

applicationsarxiv-cs-lg

When Privacy Moves ML-Mediated Decisions On Device: Information and Incentive Misalignment in Auctions

arXiv:2609.33312v1 Announce Type: cross Abstract: Moving ML-mediated decision making onto privacy-preserving clients decentralises the economic decision along with the inference. Shared budget constra

local-aiarxiv-cs-lg

WorldGraph: Graph-Native World Modeling

model-releasesarxiv-cs-cv

arXiv:2609.34159v1 Announce Type: cross Abstract: World models infer latent states of an environment to capture its underlying dynamics and predict future evolution. Many real-world environments, howe

Zero-Storage Procedural Neural Synthesis via Boundary Dynamics: Formal Verification in Lean 4 and Bare-Metal Gauntlet Validation

arXiv:2609.33066v1 Announce Type: new Abstract: Contemporary neural inference architectures rely on dense floating-point weight matrices stored in high-bandwidth memory (VRAM), incurring severe memory

local-aiarxiv-cs-lg

Acoustic-to-Text KV Compression for Full-Duplex Speech Models

arXiv:2609.31224v1 Announce Type: cross Abstract: Full-duplex speech language models continuously accumulate acoustic key-value (KV) states, making long-running interactions memory-intensive. During l

An End-to-End Pipeline for Causal ML with Continuous Treatments: An Application to Financial Decision Making

arXiv:2609.30396v1 Announce Type: cross Abstract: This paper presents an end-to-end causal machine learning (ML) pipeline designed for real-world applications with continuous treatments. The proposed

safetyarxiv-cs-lg

Asymmetric Classifier-Free Guidance for Target-Speaker ASR

arXiv:2609.30476v1 Announce Type: cross Abstract: Target-speaker automatic speech recognition (TS-ASR) must identify and transcribe a desired speaker under varying overlap and noise conditions. These

researcharxiv-cs-cl

ContraFM-S2O: Flow Matching-Based One-step SAR-to-Optical Image Translation Model with Contrastive Learning

arXiv:2609.31378v1 Announce Type: new Abstract: In recent years, diffusion models and GAN-based models have become the mainstream approaches for SAR-to-optical image translation, owing to their advant

researcharxiv-cs-cv

Agents

DualManip: Agentic Dynamic Manipulation via Dual-Path Semantic Reasoning and Geometric Adaptation

arXiv:2609.31112v1 Announce Type: new Abstract: Vision-language models (VLMs) enable open-vocabulary reasoning for robot manipulation, but their high inference latency limits responsiveness in dynamic

agentsarxiv-cs-ro

DyMD: Preserving Interaction Dynamics through Distribution Matching Distillation in Few-Step Video World Models

arXiv:2609.31349v1 Announce Type: cross Abstract: Large video diffusion models offer expressive priors for embodied prediction and learning, yet their many-step sampling remains costly for interactive

local-aiarxiv-cs-ai

Equation discovery with Bayesian tree-adjoining grammars

arXiv:2609.31368v1 Announce Type: cross Abstract: Tree-Adjoining Grammars (TAGs) have recently been introduced to Nonlinear System Identification (NLSI) as a means of encoding an entire model class as

Factorized axis convolutional gated recurrent unit with dynamic adaptive pooling for remaining useful life prediction of rolling bearings

arXiv:2609.30972v1 Announce Type: new Abstract: Convolutional neural networks (CNN) are widely used to predict the remaining useful life (RUL) of rolling bearings from time-frequency representations (

From Weak Data to Strong Policy: Q-Targets Enable Provable In-Context Reinforcement Learning

arXiv:2609.30391v1 Announce Type: new Abstract: Existing in-context reinforcement learning methods mainly pretrain Transformers with supervised behavior-prediction objectives. This enables task infere

safetyarxiv-cs-lg

Governed Deduction: Policy-Grounded Premise Authorization Beyond Relevance

model-releasesarxiv-cs-ai

arXiv:2609.31029v1 Announce Type: new Abstract: Reasoning systems usually treat premise use as a question of relevance: if a fact is available and useful, it may be selected for inference. Authorizati

GraphWrit3R: End-to-End 3D Scene Graph Writing

model-releasesarxiv-cs-cv

arXiv:2609.31595v1 Announce Type: new Abstract: 3D scene graphs provide a structured representation of complex environments by encoding objects, their semantic attributes, and the spatial and function

Guarded Gradient-Based Activation Steering of Shutdown Responses in Qwen3.5-0.8B: A Minimum-Step Policy

arXiv:2609.30326v1 Announce Type: new Abstract: Activation steering changes a model's internal activations during inference without updating its weights, but a useful intervention must determine both

HuPER: A Human-Inspired Framework for Phonetic Perception

arXiv:2602.01634v2 Announce Type: replace-cross Abstract: We propose HuPER, a human-inspired framework that models phonetic perception as adaptive inference over acoustic-phonetics evidence and lingui

Metacognitive Selective Ensemble for Mobile Systems

arXiv:2609.31031v1 Announce Type: new Abstract: Deep ensembles improve robustness in mobile sensing, but repeatedly executing many models over continuous sensor streams is costly. Selecting only a few

MM-VeriAgent: Learning to Use Extensive Tools to Verify Multimodal Misinformation with Reinforcement Learning

arXiv:2609.30698v1 Announce Type: new Abstract: Real-world multimodal misinformation often involves mixed forgery sources, requiring sample-specific detection strategies. Existing tool-augmented metho

safetyarxiv-cs-cv

MoSAR: Mixture of Semantic Attention Regimes for Learning Adaptive and Approximable Attention Geometries

model-releasesarxiv-cs-ai

arXiv:2609.31261v1 Announce Type: cross Abstract: The quadratic complexity of dense self-attention remains a central bottleneck for long-context language modeling. Many efficient alternatives address

Predicting Transmembrane Protein Topology from 3D Structure

arXiv:2609.30446v1 Announce Type: new Abstract: This paper presents a novel approach to infer protein topology using the state-of-the-art graph neural network (GNN), SchNet. The model is trained on th

Reliability-aware Cross-sample Enhancement for Robust Multimodal Sentiment Analysis

arXiv:2609.30470v1 Announce Type: new Abstract: Multimodal Sentiment Analysis (MSA) aims to infer human emotions from multiple modalities such as text, audio, and vision. In practice, inputs are often

SkillFlow: Scalable and Efficient Agent Skill Retrieval System

model-releasesarxiv-cs-ai

arXiv:2504.06188v3 Announce Type: replace Abstract: AI agents can extend their capabilities at inference time by loading reusable skills into context, yet equipping an agent with too many skills, part

Spectral Feedback for Test-Time Alignment of Protein Diffusion Models

arXiv:2609.30456v1 Announce Type: new Abstract: Reward maximization alignment methods for discrete diffusion models have primarily focused on steering the reverse process, either by influencing token

safetyarxiv-cs-ai

Strategically Diverse Sampling for Self-Training

arXiv:2609.31571v1 Announce Type: new Abstract: Many LLM training and inference methods, including RL and test-time scaling, depend on repeated sampling, but benefit only when the responses meaningful

researcharxiv-cs-cl

Applications

Towards VLA-Dreamer: Refining VLA Behavior Using World Models

arXiv:2609.31313v1 Announce Type: cross Abstract: Vision-Language-Action models (VLAs), while showing strong potential for robot control, require massive amounts of high-quality imitation learning dat

applicationsarxiv-cs-ai

Qwen3.8-Flash-Next (125B) at 12-15 tok/s on a 2021 32GB M1 Max

model-releasesr-localllama

Hi! I'm the author of MoEspresso, which is my way of putting my own ideas about inference engines to the test. A lot of the fun has been trying different design choices, measuring what happens, and fi

27 Sept 2026

Qwen3.8-Flash-Next 177B NVFP4(119GiB): SSD streaming at 9-10 tok/s on one 16 GB RTX 5060 Ti + 32 GB RAM

model-releasesr-localllama

We built an inference engine for MoE models that don't fit in VRAM + RAM. Most of the model stays on the SSD, and experts are read as tokens need them. This started as a proof of concept, and poc work

27 Sept 2026

Improved and fixed template for GPT-OSS (again). Includes preserve_thinking and fix for Unsloth-induced bug

I posted an updated GPT-OSS template a couple of months ago, which was based on Unsloth's version. It turns out that both Unsloth's version (and, thus, mine) contain a very serious bug that can degrad

local-air-localllama

26 Sept 2026

Reasoning from scratch, round number 5! This time, talking about log-probability scoring (also a great fundamental concept for loss function…

researchsebastian-raschka--x

Reasoning from scratch, round number 5! This time, talking about log-probability scoring (also a great fundamental concept for loss functions like cross-entropy in pre-training and distillation) and s

26 Sept 2026

Tutorials

Accelerate multimodal RL training with SkyRL on Amazon SageMaker HyperPod

Learn how to run SkyRL, an open-source reinforcement learning framework, on Amazon SageMaker HyperPod to post-train a Qwen3-VL-8B vision-language model with GRPO. This walkthrough covers building the

tutorialsaws-ml-blog

Accelerating Video Diffusion via Training-Free Trajectory Routing

arXiv:2609.30096v1 Announce Type: cross Abstract: Video diffusion is computationally expensive, as it requires executing a large model across many denoising steps. Even with step-distillation, inferen

safetyarxiv-cs-ai

another stealth model launch Space Bunny Alpha is now FREE on Nous Portal try with your Hermes Agent now

model-releasesnous-research--x

another stealth model launch Space Bunny Alpha is now FREE on Nous Portal try with your Hermes Agent now Meet Space Bunny Alpha on OpenRouter ⚡ A flash model with fast inference, adjustable reasoning,

BridgeMem: Causal Dyadic Transition Residuals for Temporal Knowledge Graph Forecasting

arXiv:2609.29268v1 Announce Type: new Abstract: Temporal knowledge graph forecasting aims to infer future relational facts from the temporal structure of observed events. Existing forecasters mainly s

Canopy: Exploiting Piecewise Smooth Tree Priors for Multi-Fidelity Bandits

arXiv:2609.30017v1 Announce Type: cross Abstract: Many LLM inference problems, including model routing, prefix-cache management, prompt trimming, and test-time search, can be viewed as optimization ov

local-aiarxiv-cs-ai

CounterRoute: Self-Routed Reasoning via Hierarchical Counterfactual Credit Assignment

arXiv:2609.29109v1 Announce Type: new Abstract: Reasoning-capable language models often produce long chains of thought when direct answers suffice, wasting inference compute. Many dual-mode models lea

safetyarxiv-cs-ai

FlashLoop: Fast and Memory-Efficient Looped Transformers via Lazy Updates

arXiv:2609.29812v1 Announce Type: new Abstract: Looped Transformers have attracted substantial attention as a parameter-efficient approach to increasing computational depth through repeated applicatio

HelloWorld: Towards Practical Applications of Generative Driving World Models

arXiv:2609.28931v1 Announce Type: new Abstract: Driving world models provide a promising route toward scalable counterfactual data generation and interactive simulation beyond recorded driving logs. R

researcharxiv-cs-cv

Agents

HEXIS: Compiling Skills into Extended Finite State Machines

arXiv:2609.30123v1 Announce Type: new Abstract: Agent skills provide reusable knowledge and instructions, yet agents must repeatedly infer how to apply them and which operation should follow. This cou

agentsarxiv-cs-ai

Hardware

KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization

arXiv:2609.30059v1 Announce Type: cross Abstract: Deep learning inference and training performance depends critically on GPU kernel efficiency. Modern compilers such as PyTorch Inductor automatically

hardwarearxiv-cs-ai

Tutorials

Learning New Words from Unlabeled Test Data in Automatic Speech Recognition

arXiv:2609.28877v1 Announce Type: cross Abstract: New words are invented every day. A human listener can learn a new word by hearing it clearly once and inferring its usage from sentence context. This

tutorialsarxiv-cs-cl