Search: “inference”

Opening the Black Box of Classifier-Free Guidance via Information Bottleneck

arXiv:2610.13369v1 Announce Type: new Abstract: Classifier-free guidance (CFG) is a standard technique for improving conditional generation in diffusion models, yet its activation schedule is usually

safetyarxiv-cs-lg

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Overcoming Data Imbalances with Expert Selection Biases

DGX agent

arXiv:2610.13370v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models have become the standard architecture for frontier LLMs. The multiple parallel experts in each transformer block theoret

researcharxiv-cs-cl

12 Oct 2026

Partially Informed Stochastic Optimization via Zeroth-Order Residual Tracking

arXiv:2610.13871v1 Announce Type: new Abstract: Many optimization problems provide direct access to only part of the gradient, while the remaining component must be inferred from function evaluations.

Pretraining and Test-Time Training Budgets Matter for ARC: A Controlled Study of VARC and TRM

arXiv:2610.13660v1 Announce Type: cross Abstract: The Abstraction and Reasoning Corpus (ARC) is increasingly used to evaluate whether neural networks can infer task-specific transformations from few d

researcharxiv-cs-ai

Sonic-AMG: Streaming Human Motion Generation from Spatial Audio

arXiv:2610.13300v1 Announce Type: cross Abstract: Modeling human reactions to spatial audio requires understanding both what is heard and where it comes from. Prior work generates motion offline from

v0.40.4

model-releasesollama-releases

What's Changed llama: update llama.cpp to b11521 for decision models and EmbeddingGemma 2 by @jmorganca in #18917 server: reuse verified manifests and loaded models in inference routes by @dhiltgen in

What Should Embodied Agents Remember? Benchmarking and Evaluating Memory Modality for Embodied Agents

model-releasesarxiv-cs-ro

arXiv:2610.12561v1 Announce Type: new Abstract: Memory allows large language model (LLM) agents to use past experience, often stored as textual records, when responding to requests. For embodied agent

Basalt: Flash-Next at 665 tok/s structured, 354 prose on a 5090 + 5060 Ti (2.6x Strata)

model-releasesr-localllama

Basalt: Blackwell inference engine Hi all! So over the last week I've been working on a Strata fork that's heavily tuned and can achieve throughput up to 2.6x what Strata usually does on the same weig

10 Oct 2026

Addressing Overcommitment in the Reasoning of Gendered Economic Memes under Multimodal Ambiguity

arXiv:2610.11724v1 Announce Type: new Abstract: Multimodal meme understanding is increasingly used to analyze socially sensitive content, yet existing models often exhibit biased behavior when interpr

safetyarxiv-cs-lg

DADP: Dynamic Activity-Dependent Pruning, A Reverse Hebbian-Inspired Structural Pruning Method

arXiv:2610.11853v1 Announce Type: new Abstract: Modern neural networks are heavily over-parameterized. This redundancy incurs substantial compute and memory overhead during training and inference. Exi

Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models

arXiv:2610.10859v1 Announce Type: new Abstract: Text-to-image diffusion models are increasingly distilled into few-step variants and being deployed to enable fast inference. However, their ability to

safetyarxiv-cs-cv

Hardware

GATOR: Generative and Agentic 3D Object Reconstruction From Casual Images

arXiv:2610.11215v1 Announce Type: new Abstract: Reconstructing complete, scene-aligned 3D objects from casual images requires integrating sparse, uncertain observations and inferring surfaces hidden b

hardwarearxiv-cs-cv

H2O-Lightning-4B: Apache-2.0 4B Decision model, official #1 open model on JevBench (above Jev)

model-releasesr-localllama

Disclosure: I work at H2O.ai. We released H2O-Lightning-4B, an open-weight (Apache-2.0) model for the 'decisions API' style of inference that Jev made popular: you send a state plus typed questions (p

Image Recognition with Vision and Language Embeddings of VLMs

arXiv:2509.09311v2 Announce Type: replace Abstract: Vision-language models (VLMs) have enabled strong zero-shot classification through image-text alignment. Yet, their purely visual inference capabili

safetyarxiv-cs-cv

LLM-IDEA: Identifiability-Driven Experimental Agent for Autonomous Discovery of Mechanistic World Models

model-releasesarxiv-cs-ai

arXiv:2610.11253v1 Announce Type: new Abstract: Large language model agents are being increasingly deployed as autonomous scientists, designing experiments and inferring mechanistic world models with

Agents

Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks

arXiv:2610.11794v1 Announce Type: new Abstract: Learning to act in unfamiliar environments requires agents to infer how the world works and revise that understanding as new evidence arrives. Yet limit

agentsarxiv-cs-ai

Qwen3.8-27B: 159 tok/s on R9700, 64 tok/s on Strix Halo

LemonSeed Studio is an iPad editor/IDE with on-device inference on an AMD GPU in a Thunderbolt enclosure. It embeds the unmodified upstream Linux amdgpu + amdkfd driver (mac_linuxgpu) as a PCIDriverKi

local-air-localllama

Rethinking Latency Denial-of-Service: Attacking the LLM Serving Framework, Not the Model

arXiv:2602.07878v2 Announce Type: replace-cross Abstract: LLM inference is inherently expensive, even a modest slowdown can translate into substantial operating costs and severe availability risks. Re

researcharxiv-cs-ai

S4oP: Operator-level Pruning of Structured State Space Models for Resource-Constrained Devices

model-releasesarxiv-cs-ai

arXiv:2606.18096v2 Announce Type: replace-cross Abstract: Structured State Space Models (SSMs), including the S4 and S4D architectures, have recently emerged as powerful alternatives to attention-base

Safe at One Loop, Risky at Another: Aligning Safety Across Recurrent Depths in Looped Language Models

model-releasesarxiv-cs-lg

arXiv:2610.10625v1 Announce Type: cross Abstract: Looped Language Models (LoopLMs) provide a parameter efficient approach to scaling model capabilities through repeated use of shared parameters across

Scalable AI Uncertainty Quantification via Generalized Laplace Active Subspaces

arXiv:2610.11738v1 Announce Type: new Abstract: Reliable uncertainty quantification (UQ) is essential for deploying neural networks in scientific and high-stakes applications, but full Bayesian infere

local-aiarxiv-cs-ai

Agents

TokenBank: Financial Infrastructure for AI Services

arXiv:2610.11333v1 Announce Type: new Abstract: AI services incur inference costs during execution, while revenue may arrive later. Changing API prices, limited upfront capital, and service failures c

agentsarxiv-cs-ai

Abdominal Ultrasound Simulation from Semantic Labels using Paired Label-to-Physics-Based Image Translation

arXiv:2610.08849v1 Announce Type: cross Abstract: Purpose: Current abdominal ultrasound (US) simulation methods often require CT-based anatomical references for ray-casting, limiting deformation and p

Boundary-Free Contextual Biasing: Depth-Adaptive Gating and Reading-Space Matching for Unsegmented Languages

arXiv:2610.09467v1 Announce Type: cross Abstract: Contextual biasing supplies an ASR system with a list of expected words at inference time, but existing methods rely on word boundaries that Japanese

Broadly Applicable Approximate MCMC for Switching Stochastic Differential Equations Using Uniformization and Time-Conditioned Factorized Neural Likelihood Estimation

arXiv:2610.10194v1 Announce Type: cross Abstract: Switching stochastic differential equations (SSDEs) describe continuous-time dynamics whose parameters switch according to a latent regime process tha

Consistent Distribution Matching for Data-Free Diffusion Distillation

arXiv:2610.09221v1 Announce Type: cross Abstract: Flow and diffusion models suffer from slow inference due to computationally expensive numerical integration. Distillation provides a promising way for

tutorialsarxiv-cs-cv

Despite Instructions: Frontier Agents Improvise Covert Channels at Test Time

arXiv:2609.32701v2 Announce Type: replace-cross Abstract: In security-sensitive applications, language-model agents are often required to coordinate without disclosing confidential information. Yet re

EASE: Entropy-Adaptive Distribution Shaping for Evading AI-generated Text Detectors

arXiv:2610.09976v1 Announce Type: new Abstract: AI-generated text (AIGT) detection can be sensitive to the decoding choices of the source large language model (LLM). We observe that perturbing next-to

researcharxiv-cs-cl

GeoPrior-Mamba: Structured Process Priors with Mamba for Fine-Resolution XCO2 Reconstruction

arXiv:2610.09456v1 Announce Type: new Abstract: Reconstructing fine-resolution column-averaged dry-air CO2 (XCO2) fields from sparse satellite observations requires models to infer spatial structure t

tutorialsarxiv-cs-lg

Judging in Latent Space: Efficient Generative Reward Modeling via Semantics-Preserving Compression

arXiv:2610.09788v1 Announce Type: new Abstract: Reward modeling often requires jointly representing and reasoning over multiple evaluation criteria, yet verbalizing this process token by token can inc

KVFetch: Temporal Prefetching for the Missing Half of KV Cache Compression

arXiv:2610.08811v1 Announce Type: new Abstract: As context windows scale to tens or hundreds of thousands of tokens, KV cache compression has become essential for efficient LLM inference. Existing met

Learning Unknown Constraints without Unsafe Data via Optimality and Counterfactual Regularization

model-releasesarxiv-cs-lg

arXiv:2610.09350v1 Announce Type: cross Abstract: Learning from demonstrations (LfD) provides a framework for inferring unknown constraints from locally optimal, constraint-satisfying expert behavior.

LLM4Impact: Integrating Heterogeneous Information for Scientific Impact Prediction

arXiv:2610.10138v1 Announce Type: new Abstract: Predicting the future impact of a newly published paper is challenging because it must be inferred from heterogeneous evidence available at publication

Long-WAM: Scaling the Context of World-Action Models

arXiv:2610.10528v1 Announce Type: cross Abstract: Real-time robot control demands enough visual history to infer motion and task progress, but processing that history can delay action. We present Long

local-aiarxiv-cs-cv

MeshCarve: Artisan Mesh Generation with Flow Matching in Compact Latent Spaces

arXiv:2610.09723v1 Announce Type: new Abstract: Prior artisan mesh generation works largely predict face tokens autoregressively, which makes inference slow. Recent methods instead flow match continuo

MORCA: Offline-to-Online Reinforcement Learning for Adaptive Cache Reuse in Video Diffusion Acceleration

arXiv:2610.10457v1 Announce Type: new Abstract: Diffusion Transformers (DiTs) achieve remarkable performance in video synthesis, but their iterative denoising process suffers from high inference laten

tutorialsarxiv-cs-cv

SpatialUQ: Post-Hoc Uncertainty Quantification from Spatial Consistency in Black-Box Vision Models

arXiv:2610.09498v1 Announce Type: new Abstract: Clinical vision models are often deployed as frozen black boxes with no access to internals, retraining, or ground truth at inference time. We introduce

TERRA: Learning Transportable Latent Actions through Temporal Effect Representation and Relational Alignment

arXiv:2610.09509v1 Announce Type: new Abstract: Latent actions supervise robot policies with action-like codes inferred from visual transitions, and their usefulness hinges on two questions: what a co

safetyarxiv-cs-cv

The Impact of Likelihood Tempering on the Limiting Predictive Moments of Variational Bayesian Linear Neural Networks

arXiv:2610.09132v1 Announce Type: cross Abstract: In wide Bayesian neural networks, Gaussian mean-field variational inference is prone to 'prior dominance': the Kullback-Leibler (KL) regularization te

The Long Road to the Same Answer: Cognitive Bias Under Escalating Reasoning Budgets in Large Language Models

arXiv:2610.10049v1 Announce Type: new Abstract: Reasoning models allocate extra computation at inference time and present their answers as the product of deliberate thought. If this deliberation works

Tools

Tokenomics, from the chip all the way up. @dylan522p, Founder & CEO of @SemiAnalysis_, joins the Fireworks Forge speaker list on optimizing …

Tokenomics, from the chip all the way up. @dylan522p, Founder & CEO of @SemiAnalysis_, joins the Fireworks Forge speaker list on optimizing every layer of AI infrastructure for inference at scale. Joi

toolsfireworks-ai--x

TopoCurve: Geometry-Aware Topology Reasoning via Bezier Curves in Autonomous Driving

model-releasesarxiv-cs-cv

arXiv:2610.09118v1 Announce Type: new Abstract: Topology reasoning jointly detects 3D lanes and traffic elements from multi-view images and infers their structural connectivity. Current methods model

Training Parallel Speculative Draft Models by Directly Minimizing Expected Decoding Rounds

arXiv:2610.10411v1 Announce Type: cross Abstract: Speculative decoding accelerates large language model inference by using a low-cost draft model to propose tokens that the full-size target model veri

local-aiarxiv-cs-cl

Trajectory Planning without Trajectory Data: A Manifold-Guided Approach

arXiv:2610.08863v1 Announce Type: new Abstract: A common way for trajectory planning is to leverage generative models trained on large collections of expert trajectories. At inference time, the model

tutorialsarxiv-cs-ro

Unified Multi-plane Autoregressive Diffusion for 3D Multi-contrast MRI Synthesis

arXiv:2610.09376v1 Announce Type: new Abstract: Acquiring a complete set of magnetic resonance imaging (MRI) contrasts is time-intensive and uncomfortable for patients, despite the diagnostic value of

Hardware

YANchor-4B: Effective Long-Horizon Reasoning in O(N) Time with O(1) Memory

arXiv:2610.10118v1 Announce Type: cross Abstract: Long-horizon reasoning demands access to earlier information at a manageable generation cost. Full-history attention incurs growing storage and comput

hardwarearxiv-cs-cl

YUBI-STAG: Contact and Semantic-Rich Alignment for VLAs via Automated Video-Language Grounding

arXiv:2610.09718v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models acquire broad manipulation capabilities via large-scale pretraining, yet eliciting them through language requires

local-aiarxiv-cs-cv

2x CMP 170HX 64GB: GLM-5.3-Flash at 384K context / ~90 tok/s (EXL3, HBM-first setup) + Qwen3.8 comparison