Page 103 of 130 • 1560 Total Articles

createLiveAI

Continue exploring the latest AI breakthroughs, technology insights, and industry analysis. Page 103 of our comprehensive AI news collection.

All Articles 1560 Business 249 Ethics 150 General 142 Policy 12 Research 793 Startups 13 Technology 201

📰 Latest Intelligence

Showing 12 articles on page 103 of 130

Live feed

📱 2-column layout

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

Dyna-Think: Synergizing Reasoning, Acting, and World Model Simulation in AI Agents

The paper introduces Dyna-Think, a framework that combines planning, world modeling, reasoning, and acting to improve AI agent performance in long-horizon tasks, building upon large language models like DeepSeek-R1. Through imitation learning and a two-stage training process, Dyna-Think enhances world modeling and policy performance, achieving comparable results to R1 with fewer tokens and demonstrating that better world models lead to improved reasoning and planning capabilities.

arXiv cs.AI

Ethics

📄 arXiv cs.AI

Jun 3, 2025

Ethical AI: Towards Defining a Collective Evaluation Framework

The paper proposes a modular ethical assessment framework for AI, built on interpretable ontological blocks that encode principles like fairness and accountability, aiming to enhance transparency and compliance with regulations such as the EU AI Act. This approach facilitates scalable, explainable, and auditable AI ethics, demonstrated through a real-world AI-powered investor profiling use case, though challenges in automation and probabilistic reasoning remain.

Autonomous Systems

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

Evaluation of LLMs for mathematical problem solving

This study evaluates three large language modelsGPT-4o, DeepSeek-V3, and Gemini-2.0on diverse mathematical datasets, assessing their accuracy, reasoning steps, and problem comprehension using a Structured Chain-of-Thought framework. Results indicate GPT-4o's superior stability and performance on complex problems, while each model exhibits specific strengths and weaknesses in reasoning, explanation, and logical flexibility.

GPT Google AI

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

Hidden in Plain Sight: Probing Implicit Reasoning in Multimodal Language Models

This paper analyzes how current multimodal large language models (MLLMs) handle implicit reasoning in real-world, messy environments, revealing that they often fail to detect hidden issues despite possessing relevant skills. Simple inference-time interventions, such as cautious prompting and requesting clarifications, can significantly improve their ability to identify and address implicit problems, highlighting a gap between reasoning ability and behavioral compliance.

GPT

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

MIR: Methodology Inspiration Retrieval for Scientific Research Problems

The paper introduces Methodology Inspiration Retrieval (MIR), a novel approach to retrieving prior research that can inspire solutions for new scientific problems by leveraging a Methodology Adjacency Graph (MAG) to capture methodological lineage beyond semantic similarity. Their method significantly improves retrieval performance and, combined with LLM-based re-ranking, shows promise for enhancing automated scientific discovery.

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

MIRROR: Cognitive Inner Monologue Between Conversational Turns for Persistent Reflection and Reasoning in Conversational LLMs

The MIRROR architecture enhances large language models by mimicking human inner monologue through modular reasoning and reflection, comprising a Thinker and Talker system that maintains an internal narrative for context-aware responses. Evaluated on safety-critical, multi-turn dialogues, models using MIRROR achieved up to 156% improvement in handling conflicting preferences and outperformed baseline models by 21% on average, addressing key failure modes like sycophancy and inconsistent constraint prioritization.

GPT Claude +2

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

Monitoring Robustness and Individual Fairness

Researchers propose runtime monitoring of black-box AI models to detect input-output robustness violations, such as adversarial or fairness issues, by observing sequences of model behavior and raising alarms when similar inputs yield dissimilar outputs. They introduce the tool Clemont, which employs online algorithms and data structures like binary decision diagrams to efficiently identify robustness violations in real-time, validated through benchmark studies.

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

>-

The paper highlights that, despite advances in AI, olfaction has been largely neglected due to scientific and technological challenges, despite its importance in human cognition and emotion. It advocates for increased interdisciplinary efforts to develop olfactory benchmarks and datasets, emphasizing that incorporating smell is essential for creating more embodied, ethically aligned AI systems.

research

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

Sleep Brain and Cardiac Activity Predict Cognitive Flexibility and Conceptual Reasoning Using Deep Learning

This study introduces CogPSGFormer, a multi-modal deep learning model that predicts individual cognitive performance, such as executive functions, based on sleep microstructure data from ECG and EEG signals. Evaluated on 817 participants, the model achieved 80.3% accuracy in classifying cognitive performance levels, demonstrating the potential of sleep-derived physiological signals for cognitive assessment.

Deep Learning Transformers

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

SMELLNET: A Large-scale Dataset for Real-world Smell Recognition

Researchers have developed SmellNet, a large-scale database of approximately 180,000 samples capturing diverse natural smells, to advance AI-based scent recognition. Despite promising classification accuracy, the study highlights significant technical challenges in creating robust, real-time AI smell detection systems suitable for real-world applications.

MIT News

Technology

🎓 MIT News

Jun 3, 2025

Teaching AI models what they dont know

MIT's Themis AI has developed the Capsa platform, which enhances machine-learning models by detecting and correcting unreliable outputs in real-time, thereby quantifying and addressing model uncertainty. This innovation aims to improve AI reliability in high-stakes applications across industries, preventing costly errors and increasing trust in AI systems.

Academic

arXiv cs.AI

Research

📄 arXiv cs.AI

Jun 3, 2025

The Automated but Risky Game: Modeling Agent-to-Agent Negotiations and Transactions in Consumer Markets

The paper investigates the use of AI agents in automating negotiations and transactions between consumers and merchants, highlighting significant variability in their ability to secure favorable deals. It also emphasizes the risks of fully automating deal-making, such as behavioral anomalies leading to financial losses, underscoring the need for caution in delegating business decisions to AI.

1 2 3 4 5 6 7 ... 130

Page 103 of 130 • Showing articles 1225-1236 of 1560

Quick Navigation

Jump to any page or browse by category

Latest (Page 1) Business 249 Ethics 150 General 142 Policy 12 Research 793 Startups 13