Cutoff with an O(1) window for Potts Glauber Dynamics on lattice at High Temperature
Proves an optimal-window cutoff result for high-temperature Potts Glauber dynamics, a sharp example of probability methods in statistical mechanics.
Open arXiv ↗Reading notes · papers · GitHub projects
Learn With Me
A living page for papers, repositories, and ideas I am following in mathematics, machine learning, and AI agents. I use this as a public learning trail: what looks interesting, why it matters, and what I want to understand next.
Last curated: 2026-08-28
01
Read actively
For each paper or repo, I focus on the core problem, assumptions, method, and what I can reuse.
02
Connect ideas
I group resources by themes so math, ML, and agent systems can cross-pollinate each other.
03
Build small
When a resource looks useful, the next step is a small experiment, reproduction, or note.
Probability, dynamics, optimization, and mathematical foundations
Mathematics
Papers and repositories that connect rigorous mathematical ideas with modeling, simulation, and computational tools.
Physics-Informed Neural Networks for Biot's Model via Fixed-Stress Splitting and Energy Natural Gradient Descent
Combines numerical-analysis structure with optimization geometry to make PINNs more robust on coupled poromechanics equations.
Open arXiv ↗The Birthday Paradox for non-backtracking walks on regular graphs
Resolves a fixed-degree conjecture on self-intersections of non-backtracking walks, connecting graph structure with probabilistic threshold intuition.
Open arXiv ↗stdlib-js/stdlib
A substantial JavaScript/TypeScript numerical and scientific standard library, useful for seeing how mathematical computing APIs are organized for the web ecosystem.
Open GitHub ↗root-project/root
CERN's ROOT toolkit remains a rich case study in statistical analysis, visualization, and large-scale scientific data infrastructure.
Open GitHub ↗Deep learning, evaluation, optimization, and applied ML systems
Machine Learning
A running list of ML papers and codebases I am reading to improve model-building, evaluation, and deployment intuition.
NeuronFuzz: Safety Neuron Guided Fuzzing for LLM Safety Evaluation
Uses internal safety-neuron activations as feedback for jailbreak fuzzing, a practical evaluation idea for testing aligned language models more efficiently.
Open arXiv ↗Muon with Finite Newton-Schulz: The Smoothing Benefit in Nonsmooth Nonconvex Optimization
Analyzes the finite Newton-Schulz iterations actually used by Muon, clarifying why optimizer approximations can help rather than merely introduce error.
Open arXiv ↗Diff Mining: Logit Differences Reveal Finetuning Objectives
Compares base and finetuned logits to audit what behavior a model learned, a simple lens for understanding post-training side effects.
Open arXiv ↗huggingface/transformers
The core model-definition library for modern text, vision, audio, and multimodal ML, worth tracking for practical architecture and inference patterns.
Open GitHub ↗pytorch/rl
A modular PyTorch reinforcement-learning library that exposes clean primitives for control, robotics, and multi-agent experimentation.
Open GitHub ↗Tool use, memory, autonomy, benchmarks, and agent infrastructure
AI Agents
Resources for understanding how AI agents plan, use tools, evaluate themselves, and operate as long-running systems.
Invocation-Level Reliability of Tool-Using Agents
Separates tool choice from argument correctness and shows how early invocation errors compound across multi-step agent trajectories.
Open arXiv ↗Same Model, Different Harness: Different Coding-Agent Results
Demonstrates that context handling and harness choices can materially change coding-agent benchmark outcomes even when the underlying model is fixed.
Open arXiv ↗LLM Agents for Time-Series: A Survey
Organizes time-series agents by forecasting, synthesis, anomaly diagnosis, and decision-support problems, giving a useful map of domain-specific agent design.
Open arXiv ↗trpc-group/trpc-agent-go
A Go framework for production agent systems with graph workflows, tools, memory, MCP, evaluation, and observability built in.
Open GitHub ↗OpenNSWM-Lab/FAROS
An AutoResearch runtime that orchestrates idea generation, experiments, paper writing, and review, useful for studying research-agent workflow design.
Open GitHub ↗