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.

PRELUDE: A Benchmark Designed to Require Global Comprehension and Reasoning over Long Contexts

Preacher: Paper-to-Video Agentic System

Hallucination vs interpretation: rethinking accuracy and precision in AI-assisted data extraction for knowledge synthesis

Decentralized Weather Forecasting via Distributed Machine Learning and Blockchain-Based Model Validation

Biased AI improves human decision-making but reduces trust

Personalized Feature Translation for Expression Recognition: An Efficient Source-Free Domain Adaptation Method

A Neurosymbolic Framework for Interpretable Cognitive Attack Detection in Augmented Reality

IAD-R1: Reinforcing Consistent Reasoning in Industrial Anomaly Detection

EvaDrive: Evolutionary Adversarial Policy Optimization for End-to-End Autonomous Driving

To Theoretically Understand Transformer-Based In-Context Learning for Optimizing CSMA

ASPD: Unlocking Adaptive Serial-Parallel Decoding by Exploring Intrinsic Parallelism in LLMs

BiasGym: Fantastic LLM Biases and How to Find (and Remove) Them

Yan: Foundational Interactive Video Generation

M3-Net: A Cost-Effective Graph-Free MLP-Based Model for Traffic Prediction

LLM-Driven Adaptive 6G-Ready Wireless Body Area Networks: Survey and Framework

The Illusion of Progress: Re-evaluating Hallucination Detection in LLMs

On Understanding the Dynamics of Model Capacity in Continual Learning

WeChat-YATT: A Simple, Scalable and Balanced RLHF Trainer

Improved Personalized Headline Generation via Denoising Fake Interests from Implicit Feedback

Hardness-Aware Dynamic Curriculum Learning for Robust Multimodal Emotion Recognition with Missing Modalities

Echoes of Automation: The Increasing Use of LLMs in Newsmaking

SIFThinker: Spatially-Aware Image Focus for Visual Reasoning

Shuffle-R1: Efficient RL framework for Multimodal Large Language Models via Data-centric Dynamic Shuffle

Towards Embodied Agentic AI: Review and Classification of LLM- and VLM-Driven Robot Autonomy and Interaction

Position: The Current AI Conference Model is Unsustainable! Diagnosing the Crisis of Centralized AI Conference

MSC: A Marine Wildlife Video Dataset with Grounded Segmentation and Clip-Level Captioning

Self-Questioning Language Models

Exploring the Application of Visual Question Answering (VQA) for Classroom Activity Monitoring

Oranits: Mission Assignment and Task Offloading in Open RAN-based ITS using Metaheuristic and Deep Reinforcement Learning

DeepWriter: A Fact-Grounded Multimodal Writing Assistant Based On Offline Knowledge Base

Class-Proportional Coreset Selection for Difficulty-Separable Data

Warehouse Spatial Question Answering with LLM Agent

CodeJudgeBench: Benchmarking LLM-as-a-Judge for Coding Tasks

AmpLyze: A Deep Learning Model for Predicting the Hemolytic Concentration

EXAONE Path 2.0: Pathology Foundation Model with End-to-End Supervision

GLM-4.1V-Thinking and GLM-4.5V: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning

Why Do Open-Source LLMs Struggle with Data Analysis? A Systematic Empirical Study

Discrepancy-Aware Graph Mask Auto-Encoder

Semantic Structure-Aware Generative Attacks for Enhanced Adversarial Transferability

Quantitative Comparison of Fine-Tuning Techniques for Pretrained Latent Diffusion Models in the Generation of Unseen SAR Images

PromptTSS: A Prompting-Based Approach for Interactive Multi-Granularity Time Series Segmentation

15,500 Seconds: Lean UAV Classification Using EfficientNet and Lightweight Fine-Tuning

Prompt Attacks Reveal Superficial Knowledge Removal in Unlearning Methods

Data Pruning by Information Maximization

CCL-LGS: Contrastive Codebook Learning for 3D Language Gaussian Splatting

Security Concerns for Large Language Models: A Survey

Is Quantum Optimization Ready? An Effort Towards Neural Network Compression using Adiabatic Quantum Computing

Unraveling the iterative CHAD

FreeKV: Boosting KV Cache Retrieval for Efficient LLM Inference

LaDi-WM: A Latent Diffusion-based World Model for Predictive Manipulation

Grouped Sequency-arranged Rotation: Optimizing Rotation Transformation for Quantization for Free

Adaptive Budgeted Multi-Armed Bandits for IoT with Dynamic Resource Constraints

Vision Transformers in Precision Agriculture: A Comprehensive Survey

Goal-Oriented Time-Series Forecasting: Foundation Framework Design

CAPTURe: Evaluating Spatial Reasoning in Vision Language Models via Occluded Object Counting

FinSage: A Multi-aspect RAG System for Financial Filings Question Answering

GraspClutter6D: A Large-scale Real-world Dataset for Robust Perception and Grasping in Cluttered Scenes

Hyperflux: Pruning Reveals the Importance of Weights

ToolACE-R: Model-aware Iterative Training and Adaptive Refinement for Tool Learning

UniOcc: A Unified Benchmark for Occupancy Forecasting and Prediction in Autonomous Driving

VectorFit: Adaptive Singular & Bias Vector Fine-Tuning of Pre-trained Foundation Models

BitDecoding: Unlocking Tensor Cores for Long-Context LLMs with Low-Bit KV Cache

Explainable Sentiment Analysis with DeepSeek-R1: Performance, Efficiency, and Few-Shot Learning

Continual Learning for Multiple Modalities

Advancing MAPF towards the Real World: A Scalable Multi-Agent Realistic Testbed (SMART)

LED-Merging: Mitigating Safety-Utility Conflicts in Model Merging with Location-Election-Disjoint

Boosting Cross-problem Generalization in Diffusion-Based Neural Combinatorial Solver via Inference Time Adaptation

Rhythmic sharing: A bio-inspired paradigm for zero-shot adaptive learning in neural networks

Measuring Diversity in Synthetic Datasets

Delayed Feedback Modeling with Influence Functions

Rollout Roulette: A Probabilistic Inference Approach to Inference-Time Scaling of LLMs using Particle-Based Monte Carlo Methods

CLoQ: Enhancing Fine-Tuning of Quantized LLMs via Calibrated LoRA Initialization

Fleurs-SLU: A Massively Multilingual Benchmark for Spoken Language Understanding

Interpretable Neural ODEs for Gene Regulatory Network Discovery under Perturbations

A Lightweight Transformer with Phase-Only Cross-Attention for Illumination-Invariant Biometric Authentication

Understanding Transformer-based Vision Models through Inversion

INSIGHT: Explainable Weakly-Supervised Medical Image Analysis

Visual SLAMMOT Considering Multiple Motion Models

A Training-Free Approach for Music Style Transfer with Latent Diffusion Models

Multi-objective Optimization in CPU Design Space Exploration: Attention is All You Need

DiRW: Path-Aware Digraph Learning for Heterophily

Diversifying Policy Behaviors with Extrinsic Behavioral Curiosity

Episodic Memory Verbalization using Hierarchical Representations of Life-Long Robot Experience

Neural Networks Generalize on Low Complexity Data

Knowledge-based Consistency Testing of Large Language Models

Implicit Safe Set Algorithm for Provably Safe Reinforcement Learning

An Explainable Transformer-based Model for Phishing Email Detection: A Large Language Model Approach

Communication Cost Reduction for Subgraph Counting under Local Differential Privacy via Hash Functions

Mathematical Computation and Reasoning Errors by Large Language Models

OpenCUA: Open Foundations for Computer-Use Agents

Compass-Thinker-7B Technical Report

TextQuests: How Good are LLMs at Text-Based Video Games?

On the Definition of Intelligence

Beyond Accuracy: How AI Metacognitive Sensitivity improves AI-assisted Decision Making

LAPO: Internalizing Reasoning Efficiency via Length-Adaptive Policy Optimization

FAIRGAME: a Framework for AI Agents Bias Recognition using Game Theory

MedRep: Medical Concept Representation for General Electronic Health Record Foundation Models

A Random-Key Optimizer for Combinatorial Optimization

Federated Cross-Training Learners for Robust Generalization under Data Heterogeneity

Leveraging Large Language Models for Relevance Judgments in Legal Case Retrieval

Mathematical Computation and Reasoning Errors by Large Language Models

Created by

Haebom

Author

Liang Zhang, Edith Aurora Graf

Outline

This paper presents the results of a study evaluating the accuracy of large-scale language models (LLMs), which are increasingly used for AI-based training and assessment in mathematics education. The study assessed the accuracy of solutions and inference errors at each stage for four LLMs: OpenAI GPT-4o, OpenAI o1, DeepSeek-V3, and DeepSeek-R1, solving three types of mathematical problems: arithmetic, algebra, and number theory. We intentionally created challenging problems that LLMs were prone to error, and the experiments were conducted in both single-agent and dual-agent configurations. The results showed that the OpenAI o1 model, with its enhanced reasoning capabilities, achieved the highest or near-perfect accuracy across all mathematical problem types. Error analysis revealed that procedural errors were the most frequent, significantly impacting overall performance, while conceptual errors were relatively rare. Using a dual-agent configuration significantly improved overall performance. These results provide actionable insights for improving LLM performance and highlight effective strategies for integrating LLMs into mathematics education, contributing to the accuracy of AI-based training and assessment.

Takeaways, Limitations

•

Takeaways:

◦

We demonstrate that enhanced reasoning skills play a significant role in improving the accuracy of solving mathematical problems in LLM.

◦

In the process of solving mathematics problems in LLM, procedural errors were revealed to be the main cause of errors.

◦

We show that the performance of LLM can be significantly improved by using a dual-agent configuration.

◦

Presenting actionable strategies to improve the accuracy of AI-based mathematics education and assessment.

•

Limitations:

◦

The types and number of LLMs used are limited.

◦

There may be a lack of variety in the difficulty and types of problems.

◦

Further verification of the objectivity and reliability of error analysis is needed.

◦

Further research is needed on its applicability in real-world mathematics education environments.

View PDF

Made with Slashpage