Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

HOSt3R: Keypoint-free Hand-Object 3D Reconstruction from RGB images

PediatricsMQA: a Multi-modal Pediatrics Question Answering Benchmark

MedQARo: A Large-Scale Benchmark for Medical Question Answering in Romanian

OmniCache: A Trajectory-Oriented Global Perspective on Training-Free Cache Reuse for Diffusion Transformer Models

Beyond Imaging: Vision Transformer Digital Twin Surrogates for 3D+T Biological Tissue Dynamics

TPLA: Tensor Parallel Latent Attention for Efficient Disaggregated Prefill and Decode Inference

Mitigating Hallucinations in LM-Based TTS Models via Distribution Alignment Using GFlowNets

Test-time Corpus Feedback: From Retrieval to RAG

$\Mathrm{TIME}[t]\subseteq \mathrm{SPACE}[O(\sqrt{t})]$ via Tree Height Compression

ZPD-SCA: Unveiling the Blind Spots of LLMs in Assessing Students’ Cognitive Abilities

Documenting Deployment with Fabric: A Repository of Real-World AI Governance

High-Throughput Low-Cost Segmentation of Brightfield Microscopy Live Cell Images

Fisher-Orthogonal Projection Methods for Natural Gradient Descent with Large Batches

Robust Federated Learning under Adversarial Attacks via Loss-Based Client Clustering

SupraTok: Cross-Boundary Tokenization for Enhanced Language Model Performance

Trustworthy AI Psychotherapy: Multi-Agent LLM Workflow for Counseling and Explainable Mental Disorder Diagnosis

Minimizing Surrogate Losses for Decision-Focused Learning using Differentiable Optimization

LETToT: Label-Free Evaluation of Large Language Models On Tourism Using Expert Tree-of-Thought

CURE: Critical-Token-Guided Re-Concatenation for Entropy-Collapse Prevention

Fourier-Guided Attention Upsampling for Image Super-Resolution

Fine-Grained Safety Neurons with Training-Free Continual Projection to Reduce LLM Fine Tuning Risks

Evaluating Contrast Localizer for Identifying Causal Units in Social & Mathematical Tasks in Language Models

Neural Logic Networks for Interpretable Classification

DIVER: A Multi-Stage Approach for Reasoning-intensive Information Retrieval

Diffusing the Blind Spot: Uterine MRI Synthesis with Diffusion Models

Architectural Co-Design for Zero-Shot Anomaly Detection: Decoupling Representation and Dynamically Fusing Features in CLIP

Leveraging GNN to Enhance MEF Method in Predicting ENSO

Large-scale Multi-sequence Pretraining for Generalizable MRI Analysis in Versatile Clinical Applications

SGD Convergence under Stepsize Shrinkage in Low-Precision Training

CLAP: Coreference-Linked Augmentation for Passage Retrieval

Geometry-Aware Spiking Graph Neural Network

SIFThinker: Spatially-Aware Image Focus for Visual Reasoning

4D-PreNet: A Unified Preprocessing Framework for 4D-STEM Data Analysis

MagicGUI: A Foundational Mobile GUI Agent with Scalable Data Pipeline and Reinforcement Fine-tuning

Large-Scale Model Enabled Semantic Communication Based on Robust Knowledge Distillation

Enhancing material behavior discovery using embedding-oriented Physically-Guided Neural Networks with Internal Variables

Agentic large language models improve retrieval-based radiology question answering

PARROT: An Open Multilingual Radiology Reports Dataset

Trusted Knowledge Extraction for Operations and Maintenance Intelligence

MemoryTalker: Personalized Speech-Driven 3D Facial Animation via Audio-Guided Stylization

Content-based 3D Image Retrieval and a ColBERT-inspired Re-ranking for Tumor Flagging and Staging

Controllable Hybrid Captioner for Improved Long-form Video Understanding

Combining Cost-Constrained Runtime Monitors for AI Safety

QuestA: Expanding Reasoning Capacity in LLMs via Question Augmentation

CRABS: A syntactic-semantic pincer strategy for bounding LLM interpretation of Python notebooks

FlexOlmo: Open Language Models for Flexible Data Use

Multi-Level Fusion Graph Neural Network for Molecule Property Prediction

BiMark: Unbiased Multilayer Watermarking for Large Language Models

A foundation model with multi-variate parallel attention to generate neuronal activity

Effective Red-Teaming of Policy-Adherent Agents

From Legal Texts to Defeasible Deontic Logic via LLMs: A Study in Automated Semantic Analysis

LLM-D12: A Dual-Dimensional Scale of Instrumental and Relational Dependencies on Large Language Models

AudioLens: A Closer Look at Auditory Attribute Perception of Large Audio-Language Models

Auto prompt sql: a resource-efficient architecture for text-to-sql translation in constrained environments

EPFL-Smart-Kitchen-30: Densely annotated cooking dataset with 3D kinematics to challenge video and language models

Accountability Attribution: Tracing Model Behavior to Training Processes

Equivariant Spherical Transformer for Efficient Molecular Modeling

RepoMaster: Autonomous Exploration and Understanding of GitHub Repositories for Complex Task Solving

Rethinking Gating Mechanism in Sparse MoE: Handling Arbitrary Modality Inputs with Confidence-Guided Gate

Security Concerns for Large Language Models: A Survey

Large Language Models in the Task of Automatic Validation of Text Classifier Predictions

Debate-to-Detect: Reformulating Misinformation Detection as a Real-World Debate with Large Language Models

Exploring the Vulnerability of the Content Moderation Guardrail in Large Language Models via Intent Manipulation

An Outlook on the Opportunities and Challenges of Multi-Agent AI Systems

IRONIC: Coherence-Aware Reasoning Chains for Multi-Modal Sarcasm Detection

Advancing Marine Research: UWSAM Framework and UIIS10K Dataset for Precise Underwater Instance Segmentation

Seeing the Trees for the Forest: Rethinking Weakly-Supervised Medical Visual Grounding

Traveling Across Languages: Benchmarking Cross-Lingual Consistency in Multimodal LLMs

Learning with Spike Synchrony in Spiking Neural Networks

Explainable Prediction of the Mechanical Properties of Composites with CNNs

From Unaligned to Aligned: Scaling Multilingual LLMs with Multi-Way Parallel Corpora

A Survey on the Safety and Security Threats of Computer-Using Agents: JARVIS or Ultron?

RT-Cache: Training-Free Retrieval for Real-Time Manipulation

DSADF: Thinking Fast and Slow for Decision Making

WATCH: Adaptive Monitoring for AI Deployments via Weighted-Conformal Martingales

ICQuant: Index Coding enables Low-bit LLM Quantization

Multimodal Masked Autoencoder Pre-training for 3D MRI-Based Brain Tumor Analysis with Missing Modalities

Theory of Mind in Large Language Models: Assessment and Enhancement

SVD Based Least Squares for X-Ray Pneumonia Classification Using Deep Features

VeriCoder: Enhancing LLM-Based RTL Code Generation through Functional Correctness Validation

X-Teaming: Multi-Turn Jailbreaks and Defenses with Adaptive Multi-Agents

CLaP -- State Detection from Time Series

Celler:A Genomic Language Model for Long-Tailed Single-Cell Annotation

Exponentially Weighted Instance-Aware Repeat Factor Sampling for Long-Tailed Object Detection Model Training in Unmanned Aerial Vehicles Surveillance Scenarios

ImF: Implicit Fingerprint for Large Language Models

HoarePrompt: Structural Reasoning About Program Correctness in Natural Language

More Women, Same Stereotypes: Unpacking the Gender Bias Paradox in Large Language Models

MedLoRD: A Medical Low-Resource Diffusion Model for High-Resolution 3D CT Image Synthesis

CleverDistiller: Simple and Spatially Consistent Cross-modal Distillation

AutoMisty: A Multi-Agent LLM Framework for Automated Code Generation in the Misty Social Robot

BEHAVIOR Robot Suite: Streamlining Real-World Whole-Body Manipulation for Everyday Household Activities

LiteASR: Efficient Automatic Speech Recognition with Low-Rank Approximation

Forgotten Polygons: Multimodal Large Language Models are Shape-Blind

Control Illusion: The Failure of Instruction Hierarchies in Large Language Models

Head-Specific Intervention Can Induce Misaligned AI Coordination in Large Language Models

EmoBench-M: Benchmarking Emotional Intelligence for Multimodal Large Language Models

Investigating the Robustness of Deductive Reasoning with Large Language Models

Field Matching: an Electrostatic Paradigm to Generate and Transfer Data

Evaluation of Large Language Models via Coupled Token Generation

Towards Privacy-aware Mental Health AI Models: Advances, Challenges, and Opportunities

Forgotten Polygons: Multimodal Large Language Models are Shape-Blind

Created by

Haebom

Author

William Rudman, Michal Golovanevsky, Amir Bar, Vedant Palit, Yann LeCun, Carsten Eickhoff, Ritambhara Singh

Outline

This paper addresses the mathematical problem-solving capabilities of visual-language models (MLLMs), particularly their limitations in geometric reasoning. We evaluated the geometric shape recognition and multi-step reasoning abilities of various MLLMs, and found that their accuracy in recognizing regular polygons was below 50%. We attribute this to MLLMs' reliance on intuitive associations (System 1) and their inability to utilize deliberate reasoning (System 2). This paper proposes a Visually Cued Chain-of-Thought (VC-CoT) prompting technique that explicitly references visual annotations of shapes, and improves the accuracy of GPT-4o in counting irregular polygon sides from 7% to 93%. Our conclusions highlight the importance of visually guided prompting in enhancing MLLMs' System 2 reasoning capabilities.