#Large Language Models

Google introduces TTD-DR, an innovative AI framework that mimics human research processes through iterative draft refinement, achieving top performance in complex research benchmarks.

Discover how to use Mirascope to implement the Self-Refine technique with Large Language Models, enabling iterative improvement of AI-generated responses for enhanced accuracy.

Sakana AI introduces Reinforcement-Learned Teachers (RLTs), a novel method that trains smaller models to teach reasoning to large language models efficiently using reinforcement learning focused on generating step-by-step explanations.

New research from Apple reveals why Large Language Models tend to overthink simple puzzles but struggle and give up on complex ones, highlighting challenges in AI reasoning capabilities.

'Explore the evolution, architecture, and optimization techniques for API-calling AI agents, including practical workflows and examples for engineering teams.'

AWS’s Automated Evaluation Framework offers a scalable, automated solution to improve the accuracy, fairness, and safety of Large Language Models, setting a new benchmark in AI reliability.

Microsoft and Tsinghua researchers propose Reward Reasoning Models that adaptively allocate compute resources during evaluation, significantly improving large language model judgment and alignment across complex tasks.

Ant Group introduces SEM, a reinforcement learning method that teaches large language models to decide when to use external search tools, enhancing reasoning accuracy and reducing unnecessary searches.

Tsinghua University researchers developed the Absolute Zero paradigm to train large language models without external data, using a self-evolving code executor system to enhance AI reasoning and learning.

Multi-agent AI systems are revolutionizing business operations by enabling diverse AI agents to collaborate seamlessly, enhancing efficiency and cross-departmental communication across various industries.

Researchers from Tsinghua University and Shanghai AI Lab introduce TTRL, a novel method allowing large language models to improve their performance without labeled data by leveraging self-generated pseudo-rewards during inference.