Game TheoryCooperative Game
A note on how the stag hunt, the prisoner's dilemma, and Pareto efficiency frame cooperation among learning agents.
Ideas
Memory Palace
Contemplations on everything.
Game Theory
BenchmarkingThe Science of Benchmarking
A technical note on benchmark design and evaluation quality
AI SafetyInner Alignment
Deception, reward tampering, mesa-optimization, goal misgeneralization, and why learned objectives may diverge from training objectives.
NLPSpeech Processing
A lecture note on how computers turn speech from raw sound into text, translation, and audio reasoning.
NLPHugging Face Ecosystem
A small map of the tools that make modern NLP workflows feel less mysterious.
AI SafetyAI Governance and Liability
Tort law, compute governance, export controls, China, institutional accountability, and the regulatory toolbox for governing frontier AI.
Collection · MAIA FellowshipMIT AI Safety Fundamentals
My notes and takeaways from the MIT AI Alignment (MAIA) Fellowship
Enter section ↗
AI SafetyOuter Alignment
Reward misspecification, specification gaming, RLHF, and the gap between intended objectives and operationalized training signals.
AI SafetyThreat Models
Instrumental convergence, power-seeking, bioterrorism, cyberwarfare, and gradual disempowerment as different ways AI systems could create risk.
AI SafetyTransformative AI and Current Trajectory
Scaling drivers, capability trends, and time-horizon forecasts for thinking about whether AGI-like systems may arrive sooner than institutions expect.
NLPEncoder-Decoder and T5
A note on how encoder-decoder models read one sequence and write another, from attention masks and sequence-to-sequence learning to T5 span corruption.
Political ScienceCostly Signaling Framework
A theory room for thinking about how actors communicate resolve under uncertainty, and why some signals become credible while others remain cheap talk.
NLPParsing
A note on syntactic parsing: constituency trees, context-free grammars, probabilistic parsing, CKY, and dependency relations.
NLPParameter Efficient Finetuning
A learning note on how PEFT methods technically modify or augment model behavior, with a focus on adapters, prefix tuning, LoRA, QLoRA, and distillation.
NLPModel Evaluation
A note on how I think about evaluating language models, from benchmarks and overlap metrics to confidence and the limits of automatic scoring.
NLPNeural Language Models
From one-hot vectors to embeddings, CBOW, and Skip-gram: a note on how representation learning changed NLP.
Collection · NLPNatural Language Processing
My notes and takeaways on NLP
Enter section ↗
NLPTransformers, From First Principles
A ground-up walkthrough of the transformer architecture, from tokens and embeddings to attention heads and residual streams.
NLPHypothesis Testing for Explainable Vietnamese Legal Relation Classification
Testing whether PhoBERT can classify Vietnamese legal relations robustly under wording perturbations.
Political ScienceTwo Equilibria, One Dilemma
Selective transparency versus hard censorship in Singapore and China.
AI GovernanceChina's Digital Authoritarianism: A Realist Approach
A realist reading of China’s digital authoritarianism as technology-enabled state power.
No notes found in this category yet.