Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Posts
Proofs as Observations
Published:
A question about AI and mathematics: could the proofs we obtain become the starting point for what we understand?
A Dinner with a Senior
Published:
Takeaways from dinner with a senior — on postgraduate recommendation classes, research drive, confronting real-world affairs, and starting from X_0 to X_1.
A Balance Scale with Balls
Published:
A classic puzzle about finding one abnormal ball with a balance scale.
The Problem on a Carousel
Published:
A stopping problem about placing luggage on a carousel with limited information and a hard deadline.
Choice and Alignment
Published:
How social choice theory and Arrow’s theorem help explain trade-offs in AI alignment.
From an ONNX policy to a running robot
Published:
A practical guide to running an ONNX policy on a robot with safe real-time control loops.
My Visit to Professor Ziad Obermeyer at UC Berkeley
Published:
Takeaways from meeting Professor Obermeyer at UC Berkeley School of Public Health — on applied math, interpretability, and mathematical models in healthcare.
Floating Point Notes in Computers
Published:
Notes on how computers represent floating-point numbers and why rounding errors are unavoidable.
Uncertainty and Belief
Published:
Reflections on uncertainty, exploration, and finding meaning beyond fixed paths.
CARLA Simulator on WSL2: Resolving the Docker Vulkan Trap
Published:
A guide on setting up a robust Windows-WSL2 hybrid simulation environment for CARLA 0.9.16, alongside a detailed root-cause analysis of Vulkan rendering crashes in WSL2 Docker.
When CoT Surface Features Transfer
Published:
A technical note on measurement non-invariance across math, science, and coding reasoning traces.
Intelligence Unveiled: Learning Mechanisms in Artificial and Biological Systems
Published:
Reflections on Dr. Yongping Cui’s lecture: Bridging In-Context Learning Phase Transitions with the discrete logic of animal behavior.
portfolio
Cross-Domain Limits of CoT Surface Features
A single-model measurement study showing that cheap reasoning-trace features are strong for math, partial for science, and weak for executable coding correctness.
publications
Token-Level Verification under Controlled Evaluation
Published in ICML 2026 Workshop on AI for Math (AI4Math), 2026
A controlled audit of shallow token-level verification signals for LLM math reasoning. Accepted at the ICML 2026 Workshop on AI for Math (AI4Math).
talks
From Bandits to PPO: RL Prerequisite Seminar
Published:
A one-hour mathematical bridge from bandits to PPO…
Designed as preparation for GRPO/DAPO discussions.