Search: “inference”

FASTER: Value-Guided Sampling for Fast RL

arXiv:2604.19730v1 Announce Type: cross Abstract: Some of the most performant reinforcement learning algorithms today can be prohibitively expensive as they use test-time scaling methods such as sampl

safetyarxiv-cs-ai

FlowForge: A Staged Local Rollout Engine for Flow-Field Prediction

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

arXiv:2604.18953v1 Announce Type: new Abstract: Deep learning surrogates for CFD flow-field prediction often rely on large, complex models, which can be slow and fragile when data are noisy or incompl

local-aiarxiv-cs-lg

22 Apr 2026

Flux 2-Klein-9B NVFP4 works well on my RTX 3050, but it takes 55sec to generate 1024 resolution.

local-air-stablediffusion

Flux 2-Klein-9B NVFP4 is a quantized image generation model that runs on mid-range GPUs like the RTX 3050. On this hardware, the model produces 1024-resolution images but with relatively slow inferenc

From Natural Language to Executable Narsese: A Neuro-Symbolic Benchmark and Pipeline for Reasoning with NARS

arXiv:2604.18873v1 Announce Type: new Abstract: Large language models (LLMs) are highly capable at language generation, but they remain unreliable when reasoning requires explicit symbolic structure,

GaiaFlow: Semantic-Guided Diffusion Tuning for Carbon-Frugal Search

arXiv:2602.15423v3 Announce Type: replace-cross Abstract: As the burgeoning power requirements of sophisticated neural architectures escalate, the information retrieval community has recognized ecolog

researcharxiv-cs-lg

GAIN: Multiplicative Modulation for Domain Adaptation

arXiv:2604.04516v2 Announce Type: replace-cross Abstract: Adapting LLMs to new domains causes forgetting because standard methods (e.g., full fine-tuning, LoRA) inject new directions into the weight s

Gated Memory Policy

arXiv:2604.18933v1 Announce Type: cross Abstract: Robotic manipulation tasks exhibit varying memory requirements, ranging from Markovian tasks that require no memory to non-Markovian tasks that depend

Geometric Decoupling: Diagnosing the Structural Instability of Latent

arXiv:2604.18804v1 Announce Type: cross Abstract: Latent Diffusion Models (LDMs) achieve high-fidelity synthesis but suffer from latent space brittleness, causing discontinuous semantic jumps during e

GRAFT: Geometric Refinement and Fitting Transformer for Human Scene Reconstruction

model-releasesarxiv-cs-cv

arXiv:2604.19624v1 Announce Type: new Abstract: Reconstructing physically plausible 3D human-scene interactions (HSI) from a single image currently presents a trade-off: optimization based methods off

GRASPrune: Global Gating for Budgeted Structured Pruning of Large Language Models

arXiv:2604.19398v1 Announce Type: new Abstract: Large language models (LLMs) are expensive to serve because model parameters, attention computation, and KV caches impose substantial memory and latency

HardNet++: Nonlinear Constraint Enforcement in Neural Networks

arXiv:2604.19669v1 Announce Type: new Abstract: Enforcing constraint satisfaction in neural network outputs is critical for safety, reliability, and physical fidelity in many control and decision-maki

local-aiarxiv-cs-lg

Highlights: 👉 80.2% SWE-Bench Verified and 89.6% LiveCodeBench v6 👉 Agent Swarm executes up to 4,000 coordinated steps 👉 Native text, ima…

Highlights: 👉 80.2% SWE-Bench Verified and 89.6% LiveCodeBench v6 👉 Agent Swarm executes up to 4,000 coordinated steps 👉 Native text, image, and video input with 79.4% MMMU-Pro 👉 Production-ready on t

agentstogether-ai--x

Tutorials

How Do Answer Tokens Read Reasoning Traces? Self-Reading Patterns in Thinking LLMs for Quantitative Reasoning

arXiv:2604.19149v1 Announce Type: cross Abstract: Thinking LLMs produce reasoning traces before answering. Prior activation steering work mainly targets on shaping these traces. It remains less unders

tutorialsarxiv-cs-ai

How far are we from agents that can self-generate world knowledge? The work proposes an outcome-based reward that measures how much an agent…

How far are we from agents that can self-generate world knowledge? The work proposes an outcome-based reward that measures how much an agent's self-generated world knowledge actually improves its task

model-releasesdair-ai--x

Idle is the New Sleep: Configuration-Aware Alternative to Powering Off FPGA-Based DL Accelerators During Inactivity

arXiv:2407.12027v2 Announce Type: replace-cross Abstract: In the rapidly evolving Internet of Things (IoT) domain, we concentrate on enhancing energy efficiency in Deep Learning accelerators on FPGA-b

Inductive Subgraphs as Shortcuts: Causal Disentanglement for Heterophilic Graph Learning

arXiv:2604.19186v1 Announce Type: cross Abstract: Heterophily is a prevalent property of real-world graphs and is well known to impair the performance of homophilic Graph Neural Networks (GNNs). Prior

local-aiarxiv-cs-ai

InsideOut: Measuring and Mitigating Insider-Outsider Bias in Interview Script Generation

arXiv:2509.21080v2 Announce Type: replace-cross Abstract: Advancements in Large language models (LLMs) have enabled a variety of downstream applications like story and interview script generation. How

Introducing Kimi K2.6 from @Kimi_Moonshot, a multimodal agentic model with Agent Swarm scaling to 300 sub-agents and long-horizon coding sta…

Introducing Kimi K2.6 from @Kimi_Moonshot, a multimodal agentic model with Agent Swarm scaling to 300 sub-agents and long-horizon coding stability. AI natives can now use Kimi K2.6 on Together AI and

agentstogether-ai--x

Introducing the Google Cloud Knowledge Catalog

model-releasesgoogle-cloud-ai

Traditional data catalogs were built as manual inventories for technical users, focusing on table structures rather than the deep context that AI agents need. When agents lack business semantics and d

Investigating Counterfactual Unfairness in LLMs towards Identities through Humor

arXiv:2604.18729v1 Announce Type: new Abstract: Humor holds up a mirror to social perception: what we find funny often reflects who we are and how we judge others. When language models engage with hum

safetyarxiv-cs-cl

Tutorials

IR-Flow: Bridging Discriminative and Generative Image Restoration via Rectified Flow

arXiv:2604.19680v1 Announce Type: new Abstract: In image restoration, single-step discriminative mappings often lack fine details via expectation learning, whereas generative paradigms suffer from ine

tutorialsarxiv-cs-cv

LatticeVision: Image to Image Networks for Modeling Non-Stationary Spatial Data

model-releasesarxiv-cs-lg

arXiv:2505.09803v3 Announce Type: replace-cross Abstract: In many applications, we wish to fit a parametric statistical model to a small ensemble of spatially distributed random variables ('fields').

Hardware

LBLLM: Lightweight Binarization of Large Language Models via Three-Stage Distillation

arXiv:2604.19167v1 Announce Type: cross Abstract: Deploying large language models (LLMs) in resource-constrained environments is hindered by heavy computational and memory requirements. We present LBL

hardwarearxiv-cs-ai

Learning Evolution via Optimization Knowledge Adaptation

arXiv:2501.02200v2 Announce Type: replace-cross Abstract: The iterative search process of evolutionary algorithms (EAs) encapsulates optimization knowledge within historical populations and fitness ev

LegalBench-BR: A Benchmark for Evaluating Large Language Models on Brazilian Legal Decision Classification

model-releasesarxiv-cs-cl

arXiv:2604.18878v1 Announce Type: new Abstract: We introduce LegalBench-BR, the first public benchmark for evaluating language models on Brazilian legal text classification. The dataset comprises 3,10

🦤 LeWorldModel: Learning Physics from Pixels — Stable World Models with Just Two Losses World models: 1️⃣ DINO-WM: pretrained ViT encoder (…

🦤 LeWorldModel: Learning Physics from Pixels — Stable World Models with Just Two Losses World models: 1️⃣ DINO-WM: pretrained ViT encoder (from ImageNet) → features → predictor. But encoder is frozen,

safetyyann-lecun--x

Localization-Guided Foreground Augmentation in Autonomous Driving

arXiv:2604.18940v1 Announce Type: new Abstract: Autonomous driving systems often degrade under adverse visibility conditions-such as rain, nighttime, or snow-where online scene geometry (e.g., lane di

local-aiarxiv-cs-cv

LPO: Towards Accurate GUI Agent Interaction via Location Preference Optimization

arXiv:2506.09373v3 Announce Type: replace-cross Abstract: The advent of autonomous agents is transforming interactions with Graphical User Interfaces (GUIs) by employing natural language as a powerful

agentsarxiv-cs-ai

Tutorials

MapPFN: Learning Causal Perturbation Maps in Context

arXiv:2601.21092v2 Announce Type: replace Abstract: Planning effective interventions in biological systems requires treatment-effect models that adapt to unseen biological contexts by identifying thei

tutorialsarxiv-cs-lg

MATA: Multi-Agent Framework for Reliable and Flexible Table Question Answering

arXiv:2602.09642v2 Announce Type: replace-cross Abstract: Recent advances in Large Language Models (LLMs) have significantly improved table understanding tasks such as Table Question Answering (TableQ

agentsarxiv-cs-ai

Mechanistic Anomaly Detection via Functional Attribution

model-releasesarxiv-cs-lg

arXiv:2604.18970v1 Announce Type: new Abstract: We can often verify the correctness of neural network outputs using ground truth labels, but we cannot reliably determine whether the output was produce

MedFlowSeg: Flow Matching for Medical Image Segmentation with Frequency-Aware Attention

arXiv:2604.19675v1 Announce Type: new Abstract: Flow matching has recently emerged as a principled framework for learning continuous-time transport maps, enabling efficient deterministic generation wi

researcharxiv-cs-cv

Micro Language Models Enable Instant Responses

model-releasesarxiv-cs-cl

arXiv:2604.19642v1 Announce Type: new Abstract: Edge devices such as smartwatches and smart glasses cannot continuously run even the smallest 100M-1B parameter language models due to power and compute

Mixture of Predefined Experts: Maximizing Data Usage on Vertical Federated Learning

arXiv:2602.12708v2 Announce Type: replace Abstract: Vertical Federated Learning (VFL) has emerged as a critical paradigm for collaborative model training in privacy-sensitive domains such as finance a

safetyarxiv-cs-lg

Multi-modal Test-time Adaptation via Adaptive Probabilistic Gaussian Calibration

arXiv:2604.19093v1 Announce Type: cross Abstract: Multi-modal test-time adaptation (TTA) enhances the resilience of benchmark multi-modal models against distribution shifts by leveraging the unlabeled

OmniParser V2: Structured-Points-of-Thought for Unified Visual Text Parsing and Its Generality to Multimodal Large Language Models

arXiv:2502.16161v2 Announce Type: replace-cross Abstract: Visually-situated text parsing (VsTP) has recently seen notable advancements, driven by the growing demand for automated document understandin

researcharxiv-cs-cl

On the Spatiotemporal Dynamics of Generalization in Neural Networks

arXiv:2602.01651v2 Announce Type: replace-cross Abstract: Why do neural networks fail to generalize addition from 16-digit to 32-digit numbers, while a child who learns the rule can apply it to arbitr

local-aiarxiv-cs-ai

Pause or Fabricate? Training Language Models for Grounded Reasoning

arXiv:2604.19656v1 Announce Type: new Abstract: Large language models have achieved remarkable progress on complex reasoning tasks. However, they often implicitly fabricate information when inputs are

researcharxiv-cs-cl

Personalized Embodied Navigation for Portable Object Finding

arXiv:2403.09905v5 Announce Type: replace-cross Abstract: Embodied navigation methods commonly operate in static environments with stationary objects. In this work, we present approaches for tackling

agentsarxiv-cs-cv

Plausible Reasoning and First-Order Plausible Logic

arXiv:2604.19036v1 Announce Type: new Abstract: Defeasible statements are statements that are likely, or probable, or usually true, but may occasionally be false. Plausible reasoning makes conclusions

Applications

Position: No Retroactive Cure for Infringement during Training

arXiv:2604.18649v1 Announce Type: cross Abstract: As generative AI faces intensifying legal challenges, the machine learning community has increasingly relied on post-hoc mitigation -- especially mach

applicationsarxiv-cs-ai

Applications

PREF-XAI: Preference-Based Personalized Rule Explanations of Black-Box Machine Learning Models

arXiv:2604.19684v1 Announce Type: new Abstract: Explainable artificial intelligence (XAI) has predominantly focused on generating model-centric explanations that approximate the behavior of black-box

applicationsarxiv-cs-lg

ProjLens: Unveiling the Role of Projectors in Multimodal Model Safety

arXiv:2604.19083v1 Announce Type: cross Abstract: Multimodal Large Language Models (MLLMs) have achieved remarkable success in cross-modal understanding and generation, yet their deployment is threate

safetyarxiv-cs-ai

PuzzleWorld: A Benchmark for Multimodal, Open-Ended Reasoning in Puzzlehunts

arXiv:2506.06211v2 Announce Type: replace-cross Abstract: Puzzlehunts are a genre of complex, multi-step puzzles lacking well-defined problem definitions. In contrast to conventional reasoning benchma

Q-Mask: Query-driven Causal Masks for Text Anchoring in OCR-Oriented Vision-Language Models

model-releasesarxiv-cs-cv

arXiv:2604.00161v2 Announce Type: replace Abstract: Optical Character Recognition (OCR) is increasingly regarded as a foundational capability for modern vision-language models (VLMs), enabling them no

Applications

Reduced-Order Surrogates for Forced Flexible Mesh Coastal-Ocean Models

arXiv:2602.05416v2 Announce Type: replace-cross Abstract: While proper orthogonal decomposition (POD)-based surrogates are widely explored for hydrodynamic applications, the use of Koopman autoencoder

applicationsarxiv-cs-ai

Refute-or-Promote: An Adversarial Stage-Gated Multi-Agent Review Methodology for High-Precision LLM-Assisted Defect Discovery

arXiv:2604.19049v1 Announce Type: cross Abstract: LLM-assisted defect discovery has a precision crisis: plausible-but-wrong reports overwhelm maintainers and degrade credibility for real findings. We

agentsarxiv-cs-ai

Remask, Don't Replace: Token-to-Mask Refinement in Masked Diffusion Language Models

arXiv:2604.18738v1 Announce Type: new Abstract: Masked diffusion language models such as LLaDA2.1 rely on Token-to-Token (T2T) editing to correct their own generation errors: whenever a different toke

researcharxiv-cs-cl