42 stories tagged Reinforcement Learning
Researchers introduced Behavioral Self-Rewarding (BSR), a method that lets AI agents judge their own debugging trajectories and improve performance without a separately trained reward model. BSR addresses the limitations of binary rewards in reinforcement learning with verifiable rewards (RLVR) by providing more nuanced feedback on debugging quality.
Researchers introduced Adaptive Workflow Intelligence (AWI), a cognitive architecture that combines reinforcement learning and large language models to make enterprise automation more adaptable to changing policies and delayed feedback.
Researchers introduced WM-R1, a new AI framework that trains mobile app agents using virtual worlds instead of real devices. This reduces costs and speeds up learning, potentially making AI assistants for apps more accessible.
A new arXiv study finds that AI models trained with reinforcement learning (RL) develop superior internal representations for mathematical reasoning compared to supervised fine-tuning (SFT) models, explaining their better performance on math problems.
Google has developed an AI system that uses reinforcement learning to correct errors in quantum computers, potentially making quantum computing more reliable and practical for real-world use.
A comprehensive survey on arXiv categorizes methods for verifying reinforcement learning (RL) policies, addressing a critical barrier to deploying AI in safety-critical domains like autonomous driving and healthcare.
Researchers have developed a three-stage method that translates a trained deep reinforcement learning policy into an executable Prolog program, making AI decisions readable, runnable, and editable by humans.
Researchers introduced DROPJ, a human-centered method that trains AI agents safely by combining world models with human feedback and justifications. This approach could make AI systems more reliable in safety-critical domains like healthcare and autonomous driving.
A new paper introduces a method to improve how AI agents learn from human feedback, making them better at following human values. This could lead to AI systems that are more reliable and trustworthy in real-world applications.
Researchers created Infinity-Parser2, an AI that can read and understand documents better than ever. It uses a special training method to handle all kinds of documents, from contracts to reports, in both English and Chinese. This could make document processing faster and more accurate for businesses and everyday users.
Researchers created a new system to train AI agents in realistic simulations, using reinforcement learning and reward shaping to improve multi-step decision-making.
Researchers developed a new AI approach that mimics how doctors gather evidence to make diagnoses. This method could make AI medical tools more accurate and reliable.
Researchers developed a method for AI models to reflect on past experiences and improve. This could make AI systems smarter and more adaptable over time. The approach is called procedural memory distillation.
Researchers introduced MedAgentBench-v3, a new benchmark for AI agents in healthcare. It improves on previous versions by addressing issues where agents often did nothing, making it harder to train them effectively.
Researchers developed a new approach called ATOD to improve AI agents for complex tasks. It combines imitation learning with reinforcement learning to help agents learn faster and perform better.
Researchers introduced Tree of Evidence (ToE), an AI system that builds argument trees to fact-check claims. It dynamically gathers and evaluates evidence from multiple sources to combat misinformation, including AI-generated content under Generative Engine Optimization (GEO) poisoning.
OpenAI introduced a new approach to training AI models called 'beneficial RL' — reinforcement learning designed to make AI systems more broadly and persistently beneficial. This method goes beyond standard RLHF by focusing on long-term alignment and continued helpfulness even as the environment or user needs change.
A new study explores whether reinforcement learning (RL) on beneficial behavior can help AI systems generalize alignment beyond their training data, addressing risks like reward hacking and deception in high-stakes settings.
Researchers developed an AI system that creates digital twins of patients to simulate treatments and optimize care in real-time while adhering to safety constraints. This approach could lead to more personalized and effective medical decisions.
The open-source community is rallying behind OpenEnv, a new platform for developing AI agents using reinforcement learning. This could make advanced AI tools more accessible to developers and researchers.
Researchers developed a new way to train AI agents for strategic games by delaying reward attribution so that the quality of an action can consider future events and other players' moves.
Researchers developed an AI approach that lets self-driving cars learn from experts while avoiding dangerous mistakes. The system only asks for help when it's truly uncertain, making it safer to explore new driving situations.
Researchers developed COSMO-Agent, an AI system that automates the back-and-forth process of designing and testing industrial products. This could make creating everything from cars to medical devices faster and cheaper.
Researchers developed a method called ICRL to help AI models learn from their own critiques. This could make AI assistants like ChatGPT or Claude much more reliable over time. The key is that the models internalize feedback, not just follow temporary instructions.