Search: “concepts”

arXiv:2601.14004v4 Announce Type: replace Abstract: Mechanistic Interpretability (MI) has emerged as a vital approach to demystify the opaque decision-making of Large Language Models (LLMs). However,

local-aiarxiv-cs-cl

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Local Ai

Masked by Consensus: Disentangling Privileged Knowledge in LLM Correctness

DGX agent

arXiv:2604.12373v1 Announce Type: new Abstract: Humans use introspection to evaluate their understanding through private internal states inaccessible to external observers. We investigate whether larg

local-aiarxiv-cs-cl

15 Apr 2026

MAST: Mask-Guided Attention Mass Allocation for Training-Free Multi-Style Transfer

arXiv:2604.12281v1 Announce Type: cross Abstract: Style transfer aims to render a content image with the visual characteristics of a reference style while preserving its underlying semantic layout and

On the Mathematical Relationship Between Layer Normalization and Dynamic Activation Functions

arXiv:2503.21708v4 Announce Type: replace-cross Abstract: Layer normalization (LN) is an essential component of modern neural networks. While many alternative techniques have been proposed, none of th

RoleMAG: Learning Neighbor Roles in Multimodal Graphs

arXiv:2604.12271v1 Announce Type: new Abstract: Multimodal attributed graphs (MAGs) combine multimodal node attributes with structured relations. However, existing methods usually perform shared messa

researcharxiv-cs-lg

Safe-SAIL: Towards a Fine-grained Safety Landscape of Large Language Models via Sparse Autoencoder Interpretation Framework

model-releasesarxiv-cs-ai

arXiv:2509.18127v3 Announce Type: replace-cross Abstract: Sparse autoencoders (SAEs) enable interpretability research by decomposing entangled model activations into monosemantic features. However, un

Scaling Exposes the Trigger: Input-Level Backdoor Detection in Text-to-Image Diffusion Models via Cross-Attention Scaling

arXiv:2604.12446v1 Announce Type: cross Abstract: Text-to-image (T2I) diffusion models have achieved remarkable success in image synthesis, but their reliance on large-scale data and open ecosystems i

The Enforcement and Feasibility of Hate Speech Moderation on Twitter

arXiv:2604.12289v1 Announce Type: cross Abstract: Online hate speech is associated with substantial social harms, yet it remains unclear how consistently platforms enforce hate speech policies or whet

researcharxiv-cs-cl

AI Integrity: A New Paradigm for Verifiable AI Governance

arXiv:2604.11065v1 Announce Type: new Abstract: AI systems increasingly shape high-stakes decisions in healthcare, law, defense, and education, yet existing governance paradigms -- AI Ethics, AI Safet

Beyond Statistical Co-occurrence: Unlocking Intrinsic Semantics for Tabular Data Clustering

model-releasesarxiv-cs-ai

arXiv:2604.10865v1 Announce Type: new Abstract: Deep Clustering (DC) has emerged as a powerful tool for tabular data analysis in real-world domains like finance and healthcare. However, most existing

Bridging What the Model Thinks and How It Speaks: Self-Aware Speech Language Models for Expressive Speech Generation

model-releasesarxiv-cs-cl

arXiv:2604.11424v1 Announce Type: new Abstract: Speech Language Models (SLMs) exhibit strong semantic understanding, yet their generated speech often sounds flat and fails to convey expressive intent,

Cross-Cultural Value Awareness in Large Vision-Language Models

arXiv:2604.09945v1 Announce Type: cross Abstract: The rapid adoption of large vision-language models (LVLMs) in recent years has been accompanied by growing fairness concerns due to their propensity t

Deliberative Alignment is Deep, but Uncertainty Remains: Inference time safety improvement in reasoning via attribution of unsafe behavior to base model

arXiv:2604.09665v1 Announce Type: cross Abstract: While the wide adoption of refusal training in large language models (LLMs) has showcased improvements in model safety, recent works have highlighted

Local Ai

Different types of syntactic agreement recruit the same units within large language models

arXiv:2512.03676v2 Announce Type: replace Abstract: Large language models (LLMs) can reliably distinguish grammatical from ungrammatical sentences, but how grammatical knowledge is represented within

local-aiarxiv-cs-cl

Diffusion-Based Generative Priors for Efficient Beam Alignment in Directional Networks

arXiv:2604.09653v1 Announce Type: cross Abstract: Beam alignment is a key challenge in directional mmWave and THz systems, where narrow beams require accurate yet low-overhead training. Existing learn

Applications

Domain-Specific Data Generation Framework for RAG Adaptation

arXiv:2510.11217v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) combines the language understanding and reasoning power of large language models (LLMs) with external ret

applicationsarxiv-cs-ai

Evolutionary Token-Level Prompt Optimization for Diffusion Models

arXiv:2604.09861v1 Announce Type: new Abstract: Text-to-image diffusion models exhibit strong generative performance but remain highly sensitive to prompt formulation, often requiring extensive manual

FREE-Switch: Frequency-based Dynamic LoRA Switch for Style Transfer

arXiv:2604.10023v1 Announce Type: cross Abstract: With the growing availability of open-sourced adapters trained on the same diffusion backbone for diverse scenes and objects, combining these pretrain

Latent Instruction Representation Alignment: defending against jailbreaks, backdoors and undesired knowledge in LLMs

arXiv:2604.10403v1 Announce Type: new Abstract: We address jailbreaks, backdoors, and unlearning for large language models (LLMs). Unlike prior work, which trains LLMs based on their actions when give

safetyarxiv-cs-lg

Learning Visually Interpretable Oscillator Networks for Soft Continuum Robots from Video

arXiv:2511.18322v3 Announce Type: replace-cross Abstract: Learning soft continuum robot (SCR) dynamics from video offers flexibility but existing methods lack interpretability or rely on prior assumpt

Legal2LogicICL: Improving Generalization in Transforming Legal Cases to Logical Formulas via Diverse Few-Shot Learning

arXiv:2604.11699v1 Announce Type: cross Abstract: This work aims to improve the generalization of logic-based legal reasoning systems by integrating recent advances in NLP with legal-domain adaptive f

Linear Programming for Multi-Criteria Assessment with Cardinal and Ordinal Data: A Pessimistic Virtual Gap Analysis

arXiv:2604.09555v1 Announce Type: new Abstract: Multi-criteria Analysis (MCA) is used to rank alternatives based on various criteria. Key MCA methods, such as Multiple Criteria Decision Making (MCDM)

Linguistic Accommodation Between Neurodivergent Communities on Reddit:A Communication Accommodation Theory Analysis of ADHD and Autism Groups

arXiv:2604.10063v1 Announce Type: new Abstract: Social media research on mental health has focused predominantly on detecting and diagnosing conditions at the individual level. In this work, we shift

researcharxiv-cs-cl

MM-LIMA: Less Is More for Alignment in Multi-Modal Datasets

arXiv:2308.12067v3 Announce Type: replace-cross Abstract: Multimodal large language models are typically trained in two stages: first pre-training on image-text pairs, and then fine-tuning using super

Premier: Personalized Preference Modulation with Learnable User Embedding in Text-to-Image Generation

arXiv:2603.20725v2 Announce Type: replace Abstract: Text-to-image generation has advanced rapidly, yet it still struggles to capture the nuanced user preferences. Existing approaches typically rely on

safetyarxiv-cs-cv

Prompt Injection as Role Confusion

arXiv:2603.12277v3 Announce Type: replace-cross Abstract: Language models remain vulnerable to prompt injection attacks despite extensive safety training. We trace this failure to role confusion: mode

RedNote-Vibe: A Dataset for Capturing Temporal Dynamics of AI-Generated Text in Lifestyle Social Media

arXiv:2509.22055v2 Announce Type: replace Abstract: We introduce RedNote-Vibe, a dataset spanning five years (pre-LLM to July 2025) sourced from lifestyle platform RedNote (Xiaohongshu), capturing the

researcharxiv-cs-cl

Steered LLM Activations are Non-Surjective

arXiv:2604.09839v1 Announce Type: new Abstract: Activation steering is a popular white-box control technique that modifies model activations to elicit an abstract change in output behavior. It has als

Applications

TaFall: Balance-Informed Fall Detection via Passive Thermal Sensing

arXiv:2604.09693v1 Announce Type: cross Abstract: Falls are a major cause of injury and mortality among older adults, yet most incidents occur in private indoor environments where monitoring must bala

applicationsarxiv-cs-ai

Vibe-driven model-based engineering

arXiv:2604.10645v1 Announce Type: cross Abstract: There is a pressing need for better development methods and tools to keep up with the growing demand and increasing complexity of new software systems

Why Steering Works: Toward a Unified View of Language Model Parameter Dynamics

model-releasesarxiv-cs-ai

arXiv:2602.02343v3 Announce Type: replace-cross Abstract: Methods for controlling large language models (LLMs), including local weight fine-tuning, LoRA-based adaptation, and activation-based interven

Cross-Modal Knowledge Distillation from Spatial Transcriptomics to Histology

arXiv:2604.09076v1 Announce Type: new Abstract: Spatial transcriptomics provides a molecularly rich description of tissue organization, enabling unsupervised discovery of tissue niches -- spatially co

Do LLMs Follow Their Own Rules? A Reflexive Audit of Self-Stated Safety Policies

arXiv:2604.09189v1 Announce Type: cross Abstract: LLMs internalize safety policies through RLHF, yet these policies are never formally specified and remain difficult to inspect. Existing benchmarks ev

Drift-Aware Online Dynamic Learning for Nonstationary Multivariate Time Series: Application to Sintering Quality Prediction

model-releasesarxiv-cs-lg

arXiv:2604.09358v1 Announce Type: new Abstract: Accurate prediction of nonstationary multivariate time series remains a critical challenge in complex industrial systems such as iron ore sintering. In

Enhancing the Safety of Medical Vision-Language Models by Synthetic Demonstrations

arXiv:2506.09067v2 Announce Type: replace-cross Abstract: Generative medical vision-language models~(Med-VLMs) are primarily designed to generate complex textual information~(e.g., diagnostic reports)

Explorable Theorems: Making Written Theorems Explorable by Grounding Them in Formal Representations

arXiv:2604.02598v2 Announce Type: replace-cross Abstract: LLM-generated explanations can make technical content more accessible, but there is a ceiling on what they can support interactively. Because

GRM: Utility-Aware Jailbreak Attacks on Audio LLMs via Gradient-Ratio Masking

arXiv:2604.09222v1 Announce Type: cross Abstract: Audio large language models (ALLMs) enable rich speech-text interaction, but they also introduce jailbreak vulnerabilities in the audio modality. Exis

Large Language Models Generate Harmful Content Using a Distinct, Unified Mechanism

arXiv:2604.09544v1 Announce Type: cross Abstract: Large language models (LLMs) undergo alignment training to avoid harmful behaviors, yet the resulting safeguards remain brittle: jailbreaks routinely

Learning Encodings by Maximizing State Distinguishability: Variational Quantum Error Correction

arXiv:2506.11552v2 Announce Type: replace-cross Abstract: Quantum error correction is crucial for protecting quantum information against decoherence. Traditional codes like the surface code require su

researcharxiv-cs-lg

Leave My Images Alone: Preventing Multi-Modal Large Language Models from Analyzing Images via Visual Prompt Injection

arXiv:2604.09024v1 Announce Type: cross Abstract: Multi-modal large language models (MLLMs) have emerged as powerful tools for analyzing Internet-scale image data, offering significant benefits but al

MixFlow: Mixed Source Distributions Improve Rectified Flows

arXiv:2604.09181v1 Announce Type: new Abstract: Diffusion models and their variations, such as rectified flows, generate diverse and high-quality images, but they are still hindered by slow iterative

safetyarxiv-cs-cv

Mosaic: Multimodal Jailbreak against Closed-Source VLMs via Multi-View Ensemble Optimization

arXiv:2604.09253v1 Announce Type: cross Abstract: Vision-Language Models (VLMs) are powerful but remain vulnerable to multimodal jailbreak attacks. Existing attacks mainly rely on either explicit visu

Ranked Activation Shift for Post-Hoc Out-of-Distribution Detection

arXiv:2604.08572v1 Announce Type: cross Abstract: State-of-the-art post-hoc out-of-distribution detection methods rely on intermediate layer activation editing. However, they exhibit inconsistent perf

Re-Mask and Redirect: Exploiting Denoising Irreversibility in Diffusion Language Models

arXiv:2604.08557v1 Announce Type: cross Abstract: Diffusion-based language models (dLLMs) generate text by iteratively denoising masked token sequences. We show that their safety alignment rests on a

Sentiment Classification of Gaza War Headlines: A Comparative Analysis of Large Language Models and Arabic Fine-Tuned BERT Models

model-releasesarxiv-cs-cl

arXiv:2604.08566v1 Announce Type: new Abstract: This study examines how different artificial intelligence architectures interpret sentiment in conflict-related media discourse, using the 2023 Gaza War

Silhouette Loss: Differentiable Global Structure Learning for Deep Representations

arXiv:2604.08573v1 Announce Type: cross Abstract: Learning discriminative representations is a central goal of supervised deep learning. While cross-entropy (CE) remains the dominant objective for cla

Towards Lifelong Aerial Autonomy: Geometric Memory Management for Continual Visual Place Recognition in Dynamic Environments

model-releasesarxiv-cs-cv

arXiv:2604.09038v1 Announce Type: cross Abstract: Robust geo-localization in changing environmental conditions is critical for long-term aerial autonomy. While visual place recognition (VPR) models pe

Verbalizing LLMs' assumptions to explain and control sycophancy

arXiv:2604.03058v2 Announce Type: replace-cross Abstract: LLMs can be socially sycophantic, affirming users when they ask questions like 'am I in the wrong?' rather than providing genuine assessment.