personal website and bloghttps://ibrahimahmed.ca/https://ibrahimahmed.ca/posts/fast-rl/https://ibrahimahmed.ca/posts/fast-rl/First open-source implementation of Soft Policy Optimization, an off-policy RL algorithm that works with LMs. This makes many RL experiments faster and cheaper.https://ibrahimahmed.ca/posts/real-work/https://ibrahimahmed.ca/posts/real-work/We are developing the first open-source LLM RL environment framework for real work.https://ibrahimahmed.ca/posts/vibe-coding/https://ibrahimahmed.ca/posts/vibe-coding/My intuitive and counterintuitive learningshttps://ibrahimahmed.ca/posts/self-refined-rl/https://ibrahimahmed.ca/posts/self-refined-rl/Policy gradient RL algorithms like GRPO have been used to improve LLMs' performance on verifiable tasks like math and coding problems.https://ibrahimahmed.ca/posts/warm-start-rl/https://ibrahimahmed.ca/posts/warm-start-rl/Insights from a small-scale transformers experiment.https://ibrahimahmed.ca/posts/eval-qa/https://ibrahimahmed.ca/posts/eval-qa/I used Claude to audit the grading code of 8 major LLM benchmarks and found issues throughout all of them.https://ibrahimahmed.ca/posts/fable-to-superintelligence/https://ibrahimahmed.ca/posts/fable-to-superintelligence/How real-world feedback loops could turn capable AI agents into recursively improving systems.