Skip to main content
← All Tags

Reinforcement Learning

14 articles in this category

AI NewsAI InfrastructureReinforcement Learning

NVIDIA AI Unveils ProRL Agent: Decoupled Rollout-as-a-Service for Multi-Turn LLM RL

NVIDIA’s ProRL Agent decouples rollout orchestration from training, nearly doubling Qwen3-8B performance on SWE-Bench Verified from 9.6% to 18.0%.

Read more
AI NewsReinforcement LearningAI Agents

Building DQN Agents with RLax, JAX, and Haiku: A Deep Dive into Reinforcement Learning Primitives

Learn to build a Deep Q-Learning agent from scratch using DeepMind's RLax library and JAX to solve the CartPole environment with 40,000 training frames.

Read more
AI NewsArtificial IntelligenceReinforcement Learning

Understanding Softmax Properties in Game Theory and Reinforcement Learning

An analysis of the Softmax function's mathematical properties and its critical role in multi-agent game theory and reinforcement learning algorithms.

Read more
AI NewsReinforcement LearningOffline Learning

Training Safety-Critical Reinforcement Learning Agents Offline

Conservative Q-Learning achieves a 25% higher return mean than Behavior Cloning in safety-critical environments.

Read more
AI NewsBig DataReinforcement Learning

Autonomous Spark Configuration with Reinforcement Learning

A reinforcement learning approach achieves 68.6% faster Spark execution times by autonomously learning optimal configurations.

Read more
AI NewsMedical ImagingReinforcement Learning

UniRG Achieves State-of-the-Art Medical Imaging Report Generation with Reinforcement Learning

UniRG, a new reinforcement learning framework, achieves state-of-the-art performance in medical image report generation – surpassing previous models on the ReXrank leaderboard.

Read more
AI NewsReinforcement LearningLLMs

Unlocking Agentic RL Training for GPT-OSS: A Practical Retrospective

LinkedIn successfully enabled agentic reinforcement learning training for the GPT-OSS-20B model, achieving comparable performance to OpenAI’s o3-mini and o4-mini.

Read more
AI NewsReinforcement LearningAgentic AI

Liquid AI’s LFM2-2.6B-Exp Tightens Small Model Behavior with Pure Reinforcement Learning

Liquid AI’s LFM2-2.6B-Exp surpasses DeepSeek R1-0528 on IFBench despite being 263x smaller, demonstrating efficient instruction following.

Read more
AI NewsReinforcement LearningAI Agents

Agent Lightning adds RL to AI agents without code rewrites

Agent Lightning enables reinforcement learning for AI agents, improving performance on complex tasks by up to 20% with minimal code changes.

Read more
AI NewsReinforcement LearningTechnology

Online Process Reward Learning (OPRL) Solves Sparse-Reward Mazes with Preference-Driven Shaping

OPRL solves sparse-reward mazes with 500-episode training loop achieving goal success.

Read more
AI NewsAI AgentsReinforcement Learning

Meta AI Introduces DreamGym: A Textual Experience Synthesizer For Reinforcement Learning RL Agents

Meta AI’s DreamGym achieves performance matching 80,000 real-environment interactions using solely synthetic data, scaling RL for LLM agents.

Read more
AI NewsReinforcement LearningQuantum Computing

Quantum-Inspired State Sculpting: Revolutionizing Offline Reinforcement Learning

Quantum-inspired state sculpting boosts offline RL performance with 100x fewer training samples.

Read more
AI NewsReinforcement LearningQuantum Computing

Quantum-Inspired Encoding: Revolutionizing Reinforcement Learning with Scarce Data

Quantum-inspired encoding boosts RL performance with scarce data, enabling breakthroughs in healthcare and finance.

Read more
AI NewsReinforcement LearningAlgorithm Design

Transitive RL: A Divide-and-Conquer Approach to Scalable Off-Policy Reinforcement Learning

This article introduces Transitive RL (TRL), a novel reinforcement learning algorithm that leverages a divide-and-conquer paradigm to address scalability issues in off-policy RL for long-horizon tasks.

Read more