A simplified, non-technical definition of reinforcement learning as an iterative optimization process based on external feedback.
A technical analysis of the DeepSeek model series, from V3 to the latest V3.2, covering architecture, performance, and release timeline.
A technical analysis of DeepSeek V3.2's architecture, sparse attention, and reinforcement learning updates, comparing it to other flagship AI models.
Analysis of DeepSeek V3.2's architecture, sparse attention mechanism, and RL updates compared to its predecessor and proprietary models.
A technical lecture on applying policy gradient methods to derive optimization algorithms, focusing on the unbiased gradient estimator and its applications.
Explores Andrej Karpathy's concept of Software 2.0, where AI writes programs through objectives and gradient descent, focusing on task verifiability.
Explores the shift from RLHF to RLVR for training LLMs, focusing on using objective, verifiable rewards to improve reasoning and accuracy.
A curated list of key LLM research papers from Jan-June 2025, organized by topic including reasoning models, RL methods, and efficient training.
A curated list of key LLM research papers from the first half of 2025, organized by topic such as reasoning models and reinforcement learning.
An AI engineer explains how buying a printer for reading and annotating technical papers helps improve focus and retention.
Analyzes the use of reinforcement learning to enhance reasoning capabilities in large language models (LLMs) like GPT-4.5 and o3.
Explores the latest developments in using reinforcement learning to improve reasoning capabilities in large language models (LLMs).
An introduction to reasoning in Large Language Models, covering concepts like chain-of-thought and methods to improve LLM reasoning abilities.
Explores four main approaches to building and enhancing reasoning capabilities in Large Language Models (LLMs) for complex tasks.
A technical guide on fine-tuning IBM's Granite3.1 AI model using Guided Reward Policy Optimization (GRPO) to enhance its reasoning capabilities.
A tutorial on reproducing DeepSeek R1's RL 'aha moment' using Group Relative Policy Optimization (GRPO) to train a model on the Countdown numbers game.
Explains the training of DeepSeek-R1, focusing on the Group Relative Policy Optimization (GRPO) reinforcement learning method.
Explores reward hacking in reinforcement learning, where AI agents exploit reward function flaws, and its critical impact on RLHF and language model alignment.
A technical review of April 2024's major open LLM releases (Mixtral, Llama 3, Phi-3, OpenELM) and a comparison of DPO vs PPO for LLM alignment.
A review and comparison of the latest open LLMs (Mixtral, Llama 3, Phi-3, OpenELM) and a study on DPO vs. PPO for LLM alignment.
