Search: “models”

Optimizing LLM Inference: Fluid-Guided Online Scheduling with Memory Constraints

arXiv:2504.11320v3 Announce Type: replace-cross Abstract: Large language models now serve millions of users daily, with providers incurring costs exceeding $700,000 per day. Each request requires toke

18 May 2026

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

PerfCodeBench: Benchmarking LLMs for System-Level High-Performance Code Optimization

DGX agent

arXiv:2605.15222v1 Announce Type: cross Abstract: Large language models (LLMs) can often generate functionally correct code, but their ability to produce efficient implementations for performance-crit

model-releasesarxiv-cs-cl

18 May 2026

Representation Without Reward: A JEPA Audit for LLM Fine-Tuning

arXiv:2605.15394v1 Announce Type: cross Abstract: Joint-embedding predictive architectures (JEPAs) propose that a model should learn more useful abstractions when trained to predict latent representat

18 May 2026

SkillSmith: Compiling Agent Skills into Boundary-Guided Runtime Interfaces

arXiv:2605.15215v1 Announce Type: new Abstract: Recently, skills have been widely adopted in large language model (LLM)-based agent systems across various domains. In existing frameworks, skills are t

18 May 2026

Cattle Trade: A Multi-Agent Benchmark for LLM Bluffing, Bidding, and Bargaining

arXiv:2605.14537v1 Announce Type: new Abstract: We introduce extsc{Cattle Trade, a multi-agent benchmark for evaluating large language models (LLMs) as agents in strategic reasoning under imperfect in

ClawForge: Generating Executable Interactive Benchmarks for Command-Line Agents

arXiv:2605.14133v1 Announce Type: new Abstract: Interactive agent benchmarks face a tension between scalable construction and realistic workflow evaluation. Hand-authored tasks are expensive to extend

Communication-Efficient Federated Fine-Tuning

arXiv:2505.04535v3 Announce Type: replace Abstract: Federated Learning (FL) enables the utilization of vast, previously inaccessible data sources. At the same time, pre-trained Language Models (LMs) h

Descriptor: Distance-Annotated Traffic Perception Question Answering (DTPQA)

arXiv:2511.13397v2 Announce Type: replace-cross Abstract: The remarkable progress of Vision-Language Models (VLMs) on a variety of tasks has raised interest in their application to automated driving.

Do-Undo Bench: Reversibility for Action Understanding in Image Generation

arXiv:2512.13609v2 Announce Type: replace Abstract: We introduce the Do-Undo task and benchmark to address a critical gap in vision-language models: understanding and generating plausible scene transf

EndPrompt: Efficient Long-Context Extension via Terminal Anchoring

arXiv:2605.14589v1 Announce Type: new Abstract: Extending the context window of large language models typically requires training on sequences at the target length, incurring quadratic memory and comp

ExploitBench: A Capability Ladder Benchmark for LLM Cybersecurity Agents

arXiv:2605.14153v1 Announce Type: cross Abstract: Exploitation is not a binary event. It is a ladder of acquiring progressive capabilities, from executing a single buggy line of code to taking full co

Forgetting That Sticks: Quantization-Permanent Unlearning via Circuit Attribution

arXiv:2605.15138v1 Announce Type: cross Abstract: Standard unlearning evaluations measure behavioral suppression in full precision, immediately after training, despite every deployed language model be

HDRFace: Rethinking Face Restoration with High-Dimensional Representation

arXiv:2605.14821v1 Announce Type: new Abstract: Face restoration under complex degradations still remains an ill-posed inverse problem due to severe information loss. Although diffusion models benefit

HiSem: Hierarchical Semantic Disentangling for Remote Sensing Image Change Captioning

arXiv:2605.15024v1 Announce Type: new Abstract: Remote sensing image change captioning (RSICC) aims to achieve high-level semantic understanding of genuine changes occurring between bi-temporal images

LLMs Know When They Know, but Do Not Act on It: A Metacognitive Harness for Test-time Scaling

arXiv:2605.14186v1 Announce Type: new Abstract: Large language models (LLMs) often expose useful signals of self-monitoring: before solving a problem, they can estimate whether they are likely to succ

NodeSynth: Socially Aligned Synthetic Data for AI Evaluation

arXiv:2605.14381v1 Announce Type: cross Abstract: Recent advancements in generative AI facilitate large-scale synthetic data generation for model evaluation. However, without targeted approaches, thes

Test-Time Learning with an Evolving Library

arXiv:2605.14477v1 Announce Type: new Abstract: We introduce EvoLib, a test-time learning framework that enables large language models to accumulate, reuse, and evolve knowledge across problem instanc

Text Knows What, Tables Know When: Clinical Timeline Reconstruction via Retrieval-Augmented Multimodal Alignment

arXiv:2605.15168v1 Announce Type: cross Abstract: Reconstructing precise clinical timelines is essential for modeling patient trajectories and forecasting risk in complex, heterogeneous conditions lik

TFGN: Task-Free, Replay-Free Continual Pre-Training Without Catastrophic Forgetting at LLM Scale

arXiv:2605.15053v1 Announce Type: cross Abstract: Continually pre-training a large language model on heterogeneous text domains, without replay or task labels, has remained an unsolved architectural p

Applications

What Makes Words Hard? Sakura at BEA 2026 Shared Task on Vocabulary Difficulty Prediction

arXiv:2605.14257v1 Announce Type: new Abstract: We describe two types of models for vocabulary difficulty prediction: a high-accuracy black-box model, which achieved the top shared task result in the

applicationsarxiv-cs-cl

Tutorials

Characteristic Root Analysis and Regularization for Linear Time Series Forecasting

arXiv:2509.23597v5 Announce Type: replace-cross Abstract: Time series forecasting remains a critical challenge across numerous domains, yet the effectiveness of complex models often varies unpredictab

tutorialsarxiv-cs-ai

Controlling Logical Collapse in LLMs via Algebraic Ontology Projection over F2

arXiv:2605.12968v1 Announce Type: cross Abstract: Do large language models internally encode ontological relations in a formally verifiable algebraic structure? We introduce Algebraic Ontology Project

CR-Net: Scaling Parameter-Efficient Training with Cross-Layer Low-Rank Structure

arXiv:2509.18993v3 Announce Type: replace Abstract: Low-rank architectures have become increasingly important for efficient large language model (LLM) pre-training, providing substantial reductions in

DocAtlas: Multilingual Document Understanding Across 80+ Languages

arXiv:2605.12623v1 Announce Type: cross Abstract: Multilingual document understanding remains limited for low-resource languages due to scarce training data and model-based annotation pipelines that p

Efficient compression of neural networks and datasets

arXiv:2505.17469v2 Announce Type: replace-cross Abstract: Compression and generalization are fundamentally related through Solomonoff induction and the minimum description length principle (MDL), whic

Tutorials

How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability

arXiv:2601.19208v2 Announce Type: replace-cross Abstract: Semantic associations such as the link between 'bird' and 'flew' are foundational for language modeling as they enable models to go beyond mem

tutorialsarxiv-cs-lg

Identifying the nonlinear string dynamics with port-Hamiltonian neural networks

arXiv:2605.12785v1 Announce Type: new Abstract: Hybrid machine learning combines physical knowledge with data-driven models to enhance interpretability and performance. In this context, Port-Hamiltoni

LoRA-Mixer: Coordinate Modular LoRA Experts Through Serial Attention Routing

arXiv:2507.00029v2 Announce Type: replace-cross Abstract: Recent attempts to combine low-rank adaptation (LoRA) with mixture-of-experts (MoE) for multi-task adaptation of Large Language Models (LLMs)

Many-Shot CoT-ICL: Making In-Context Learning Truly Learn

arXiv:2605.13511v1 Announce Type: cross Abstract: In-context learning (ICL) adapts large language models (LLMs) to new tasks by conditioning on demonstrations in the prompt without parameter updates.

MindVLA-U1: VLA Beats VA with Unified Streaming Architecture for Autonomous Driving

arXiv:2605.12624v1 Announce Type: cross Abstract: Autonomous driving has progressed from modular pipelines toward end-to-end unification, and Vision-Language-Action (VLA) models are a natural extensio

PanoWorld: Towards Spatial Supersensing in 360^irc Panorama World

arXiv:2605.13169v1 Announce Type: cross Abstract: Multimodal large laboratory models (MLLMs) still struggle with spatial understanding under the dominant perspective-image paradigm, which inherits the

Reasoning to Edit: Hypothetical Instruction-Based Image Editing with Visual Reasoning

arXiv:2507.01908v3 Announce Type: replace Abstract: Instruction-based image editing (IIE) has advanced rapidly with the success of diffusion models. However, existing efforts primarily focus on simple

Applications

Representing Higher-Order Networks: A Survey of Graph-Based Frameworks

arXiv:2605.12509v1 Announce Type: cross Abstract: Many real-world phenomena are naturally modeled by graphs and networks. However, classical graph models are often limited to pairwise interactions and

applicationsarxiv-cs-ai

Seg-Agent: Test-Time Multimodal Reasoning for Training-Free Language-Guided Segmentation

arXiv:2605.12953v1 Announce Type: cross Abstract: Language-guided segmentation transcends the scope limitations of traditional semantic segmentation, enabling models to segment arbitrary target region

Tighter Learning Guarantees on Digital Computers via Concentration of Measure on Finite Spaces

arXiv:2402.05576v4 Announce Type: replace Abstract: Machine learning models with inputs in a Euclidean space R^d, when implemented on digital computers, generalize, and their generalization gap conver

Vividh-ASR: A Complexity-Tiered Benchmark and Optimization Dynamics for Robust Indic Speech Recognition

arXiv:2605.13087v1 Announce Type: cross Abstract: Fine-tuning multilingual ASR models like Whisper for low-resource languages often improves read speech but degrades spontaneous audio performance, a p

CAD-feature enhanced machine learning for manufacturing effort estimation on sheet metal bending parts

arXiv:2605.12266v1 Announce Type: new Abstract: Graph-based machine learning has emerged as a promising approach for manufacturability analysis by learning directly from CAD models represented as Boun

Safety

Can Graphs Help Vision SSMs See Better?

arXiv:2605.11300v1 Announce Type: new Abstract: Vision state space models inherit the efficiency and long-range modeling ability of Mamba-style selective scans. However, their performance depends crit

safetyarxiv-cs-cv

Crash Assessment via Mesh-Based Graph Neural Networks and Physics-Aware Attention

arXiv:2605.11784v1 Announce Type: cross Abstract: Full-vehicle crash simulations are computationally expensive, limiting their use in iterative design exploration. This work investigates learned hybri

Keeping Score: Efficiency Improvements in Neural Likelihood Surrogate Training via Score-Augmented Loss Functions

arXiv:2605.12118v1 Announce Type: cross Abstract: For stochastic process models, parameter inference is often severely bottlenecked by computationally expensive likelihood functions. Simulation-based

LatentHDR: Decoupling Exposure from Diffusion via Conditional Latent-to-Latent Mapping for Text/Image-to-Panoramic HDR

arXiv:2605.11115v1 Announce Type: new Abstract: High Dynamic Range (HDR) generation remains challenging for generative models, which are largely limited to low dynamic range outputs. Recent diffusionb

Learning to Foresee: Unveiling the Unlocking Efficiency of On-Policy Distillation

arXiv:2605.11739v1 Announce Type: new Abstract: On-policy distillation (OPD) has emerged as an efficient post-training paradigm for large language models. However, existing studies largely attribute t

Long Story Short: Disentangling Compositionality and Long-Caption Understanding in Contrastive VLMs

arXiv:2509.19207v2 Announce Type: replace Abstract: Contrastive vision-language models (VLMs) have made significant progress in binding visual and textual information, yet understanding long, composit

Measuring Five-Nines Reliability: Sample-Efficient LLM Evaluation in Saturated Benchmarks

arXiv:2605.11209v1 Announce Type: new Abstract: While existing benchmarks demonstrate the near-perfect performance of large language models (LLMs) on various tasks, this apparent saturation often obsc

MedHopQA: A Disease-Centered Multi-Hop Reasoning Benchmark and Evaluation Framework for LLM-Based Biomedical Question Answering

arXiv:2605.12361v1 Announce Type: new Abstract: Evaluating large language models (LLMs) in the biomedical domain requires benchmarks that can distinguish reasoning from pattern matching and remain dis

Research

On Predicting the Post-training Potential of Pre-trained LLMs

arXiv:2605.11978v1 Announce Type: new Abstract: The performance of Large Language Models (LLMs) on downstream tasks is fundamentally constrained by the capabilities acquired during pre-training. Howev

researcharxiv-cs-cl

Overtrained, Not Misaligned

arXiv:2605.12199v1 Announce Type: new Abstract: Emergent misalignment (EM), where fine-tuning on a narrow task (like insecure code) causes broad misalignment across unrelated domains, was first demons

PreScam: A Benchmark for Predicting Scam Progression from Early Conversations

arXiv:2605.12243v1 Announce Type: new Abstract: Conversational scams, such as romance and investment scams, are emerging as a major form of online fraud. Unlike one-shot scam lures such as fake lotter