The Mathematics Search Engine

Mathematics News & Resources

4Mathematics is a specialist search engine for Mathematics. Discover the latest math news and mathematical content. Part of the 4SEARCH network of topic specific search engines.

Latest News & Web Pages

Quantum Zeitgeist
quantumzeitgeist.com > duet-achieves-efficient-llm-unlearning

Duet Achieves Efficient LLM Unlearning, Avoiding Catastrophic Forgetting And Reverse Attacks

Duet Achieves Efficient LLM Unlearning, Avoiding Catastrophic Forgetting And Reverse Attacks1+ hour, 17+ min ago   (1095+ words) Researchers have developed a new unlearning technique, DUET, which efficiently removes unwanted information from large language models by distilling knowledge from a specially guided "teacher" model, thereby improving both accuracy and retention of useful information with significantly less data than…...

thehansindia.com
thehansindia.com > news > national > ai-led-skilling-for-a-technology-driven-economy-1044431

AI-led skilling for a technology-driven economy

AI-led skilling for a technology-driven economy1+ hour, 24+ min ago   (124+ words) Underscores the urgency of reorienting India's skilling ecosystem toward adaptability and AI-driven learning. Prof. Saurabh Mittal, Chair Executive, FORE School of Management, New Delhi "The... Underscores the urgency of reorienting India's skilling ecosystem toward adaptability and AI-driven learning Saurabh Mittal,…...

lesswrong.com
lesswrong.com > posts > xCtBpJwpkkbkfmApc > do-llms-learn-our-preferences-or-just-our-behaviors

Do LLMs Learn Our Preferences or Just Our Behaviors? — LessWrong

Do LLMs Learn Our Preferences or Just Our Behaviors? — LessWrong1+ hour, 42+ min ago   (145+ words) There's a paper empirically measuring this that not many people here seem to have read. Ashkinaze et al. created training data where moral values were confounded with surface features. Kindness always expressed formally, fairness always expressed casually. Then they broke…...

Quantum Zeitgeist
quantumzeitgeist.com > 90-percent-accuracy-evaluation-llm-achieves-define-test-diagnose

LLM Evaluation Achieves 90% Accuracy With New Define-Test-Diagnose-Fix Workflow

LLM Evaluation Achieves 90% Accuracy With New Define-Test-Diagnose-Fix Workflow2+ hour, 13+ min ago   (785+ words) Researchers have developed a practical, iterative testing workflow and a tiered evaluation suite to reliably assess the performance of large language models, revealing that improvements in one area can inadvertently diminish accuracy in others, necessitating careful calibration of claims rather…...

BIOENGINEER.ORG
bioengineer.org > enhancing-ai-with-trusted-third-party-federated-feature-selection

Enhancing AI with Trusted Third-Party Federated Feature Selection

Enhancing AI with Trusted Third-Party Federated Feature Selection2+ hour, 14+ min ago   (1097+ words) In the evolving landscape of artificial intelligence, one of the most profound challenges faced by researchers and developers alike is the selection and optimization of features for machine learning algorithms. The ability to identify which attributes most significantly influence outcomes…...

BIOENGINEER.ORG
bioengineer.org > trusted-third-party-boosts-federated-swarm-feature-selection

Trusted Third-Party Boosts Federated Swarm Feature Selection

Trusted Third-Party Boosts Federated Swarm Feature Selection2+ hour, 20+ min ago   (1166+ words) In an era defined by unprecedented data growth and the evolution of artificial intelligence, researchers are tirelessly seeking methods to optimize the way we analyze and interpret this vast pool of information. One pioneering approach that has emerged recently is…...

Quantum Zeitgeist
quantumzeitgeist.com > proactive-interactive-reasoning-pir-advances

Proactive Interactive Reasoning (PIR) Advances LLM Reasoning By Interleaving 1 Step Of Clarification

Proactive Interactive Reasoning (PIR) Advances LLM Reasoning By Interleaving 1 Step Of Clarification2+ hour, 22+ min ago   (1210+ words) Researchers have developed a new artificial intelligence technique, Proactive Interactive Reasoning, which improves accuracy in complex tasks by enabling large language models to ask clarifying questions rather than attempting to reason with incomplete or ambiguous information. https://quantumzeitgeist.com/wp-content/uploads/Image_fx-27-24.jpg Scientists are tackling…...

Quantum Zeitgeist
quantumzeitgeist.com > reinforcement-training-dynaweb-advances-web-agent-model

Dynaweb Advances Web Agent Training Via Model-Based Reinforcement Learning

Dynaweb Advances Web Agent Training Via Model-Based Reinforcement Learning2+ hour, 33+ min ago   (915+ words) Researchers have created a new framework, DynaWeb, that trains artificial intelligence web agents using a simulated internet environment , allowing them to "imagine" interactions and learn more efficiently than previously possible. https://quantumzeitgeist.com/wp-content/uploads/Image_fx-25-24.jpg Researchers are tackling the difficult problem of creating truly…...

DEV Community
dev.to > sectorhqco > sector-hq-weekly-digest-february-1-2026-53og

Sector HQ Weekly Digest - February 1, 2026

Sector HQ Weekly Digest - February 1, 20263+ hour, 10+ min ago   (91+ words) Who's shipping vs who's just talking? Here's this week's AI industry intelligence. No high hype alerts this week The AI industry continues to evolve rapidly. Companies that ship consistently rise in our rankings, while those focused on hype alone get…...

lesswrong.com
lesswrong.com > posts > q8ZWCKZd24wDrGX7p > how-does-reasoning-affect-ethical-moral-task-results

How does reasoning affect Ethical/Moral task results? — LessWrong

8+ hour, 21+ min ago   (521+ words) So I tried evaluating how a model may perform on Moral/ethical reasoning tasks as reasoning increases. In order to elicit a reasoning increase, I utilized 4 different prompt styles: Aside: In figures it will show prompts 0, 2, 4, 5. For clarity these are…...