# Topic pool for the embedding-default eval corpus.
#
# One canonical English Wikipedia title per line. `#` comments + blank lines
# are ignored. build-corpus.mjs resolves each title's ar / he / ru / zh / ja
# counterpart via the langlinks API, so the SAME semantic topic set is used in
# every language slice — language/script is the only variable across slices.
#
# The pool is OVER-PROVISIONED on purpose. A topic is only usable if it has an
# article clearing MIN_DOC_CHARS in EVERY one of the six languages, and Hebrew
# + Arabic Wikipedia carry many more stubs than English does. The pool is sized
# so ~200 topics survive that six-way intersection; the survivors are then
# capped to exactly 200 (see TARGET_DOCS in build-corpus.mjs). Topics that fail
# in one language are reported in the manifest's `no_article` / `dropped`
# fields — they are documented, never silently substituted.
#
# Selection rule (documented so a reviewer can judge difficulty): the pool is
# deliberately CLUSTERED inside gbrain's own domain (AI research, AI labs, AI
# researchers, retrieval/IR, datastores, venture funding). Near-neighbour
# clusters make retrieval non-trivial — a pool of 200 unrelated topics would
# ceiling out at nDCG ~1.0 for every model and measure nothing.

# --- AI labs / companies (near-neighbour cluster) ---
OpenAI
Anthropic
Google DeepMind
Mistral AI
Hugging Face
Nvidia
Meta Platforms
Stability AI
Databricks
Palantir Technologies
Microsoft
Apple Inc.
Amazon (company)
Alphabet Inc.
Google
IBM
Intel
Advanced Micro Devices
Qualcomm
TSMC
Arm Holdings
Salesforce
Oracle Corporation
SAP
Adobe Inc.
Tesla, Inc.
SpaceX
Uber
Airbnb
Shopify
Spotify
Netflix
Twitter
LinkedIn
GitHub
GitLab
Dropbox
Atlassian
Cloudflare
Samsung Electronics
Sony
Alibaba Group
Tencent
Baidu
ByteDance

# --- AI researchers / founders / computing figures (near-neighbour cluster) ---
Geoffrey Hinton
Yann LeCun
Yoshua Bengio
Ilya Sutskever
Demis Hassabis
Fei-Fei Li
Sam Altman
Jensen Huang
Dario Amodei
Andrew Ng
Judea Pearl
Marvin Minsky
John McCarthy (computer scientist)
Herbert A. Simon
Alan Turing
John von Neumann
Claude Shannon
Ada Lovelace
Grace Hopper
Donald Knuth
Edsger W. Dijkstra
Tim Berners-Lee
Linus Torvalds
Guido van Rossum
Ken Thompson
Dennis Ritchie
Bjarne Stroustrup
Alan Kay
Vint Cerf
Barbara Liskov
Leslie Lamport
Richard Stallman
Steve Jobs
Bill Gates
Elon Musk
Jeff Bezos
Mark Zuckerberg
Larry Page
Sergey Brin
Satya Nadella
Sundar Pichai
Peter Thiel
Marc Andreessen

# --- Model architectures + training concepts (near-neighbour cluster) ---
Transformer (deep learning architecture)
BERT (language model)
GPT-4
GPT-3
GPT-2
Generative pre-trained transformer
Word embedding
Word2vec
Large language model
Retrieval-augmented generation
Convolutional neural network
Recurrent neural network
Long short-term memory
Diffusion model
Generative adversarial network
Autoencoder
Variational autoencoder
Attention (machine learning)
Vision transformer
Mixture of experts
Gradient descent
Stochastic gradient descent
Backpropagation
Overfitting
Regularization (mathematics)
Batch normalization
Loss function
Activation function
Softmax function
Sigmoid function
Rectifier (neural networks)
Perceptron
Multilayer perceptron
Neural network (machine learning)
Deep learning
Machine learning
Artificial intelligence
Reinforcement learning
Supervised learning
Unsupervised learning
Self-supervised learning
Transfer learning
Reinforcement learning from human feedback
Fine-tuning (deep learning)
Prompt engineering
Hallucination (artificial intelligence)
Foundation model
Multimodal learning
Byte pair encoding
Perplexity
Curse of dimensionality
Bias–variance tradeoff
Cross-validation (statistics)
Feature engineering
Precision and recall
Confusion matrix
Receiver operating characteristic
F-score
Artificial general intelligence
AI alignment
Ethics of artificial intelligence
Technological singularity

# --- Named model products (very-near-neighbour cluster: all chat LLMs) ---
ChatGPT
Claude (AI)
Google Gemini
Llama (language model)
Chatbot
Stable Diffusion
DALL-E
Midjourney
AlphaGo
AlphaFold
AlphaZero
Deep Blue (chess computer)
Virtual assistant
Siri
Amazon Alexa
Google Assistant

# --- Retrieval / IR / ranking (near-neighbour cluster — gbrain's core domain) ---
Cosine similarity
K-nearest neighbors algorithm
Search engine
Tf–idf
Recommender system
Collaborative filtering
Cluster analysis
Support vector machine
Random forest
Decision tree learning
Gradient boosting
XGBoost
AdaBoost
Naive Bayes classifier
Logistic regression
Linear regression
K-means clustering
Hierarchical clustering
DBSCAN
Principal component analysis
Singular value decomposition
Latent semantic analysis
Latent Dirichlet allocation
Topic model
Regression analysis
Bayes' theorem
Data mining
Text mining
Information retrieval
Vector space model
Okapi BM25
Inverted index
PageRank
Web crawler
Query expansion
Relevance feedback
Learning to rank
Nearest neighbor search
Locality-sensitive hashing
Full-text search
Apache Lucene
Apache Solr
Levenshtein distance
Jaccard index
Euclidean distance
Dot product
Hamming distance

# --- NLP / perception tasks ---
Natural language processing
Machine translation
Speech recognition
Speech synthesis
Computer vision
Turing test
Sentiment analysis
Named-entity recognition
Part-of-speech tagging
Word-sense disambiguation
Automatic summarization
Question answering
Stemming
Lemmatisation
Stop word
N-gram
Bag-of-words model
Optical character recognition
Digital image processing
Computer graphics
Ray tracing (graphics)
Robotics
Self-driving car

# --- Datastores + infra (near-neighbour cluster) ---
PostgreSQL
SQLite
Redis
MySQL
MariaDB
Microsoft SQL Server
Oracle Database
MongoDB
Apache Cassandra
Neo4j
ClickHouse
Memcached
Elasticsearch
Apache Kafka
RabbitMQ
Apache Spark
Apache Hadoop
Kubernetes
Docker (software)
NoSQL
SQL
Relational database
Graph database
Document-oriented database
Data warehouse
Extract, transform, load
Database transaction
Database index
Database normalization
Replication (computing)
CAP theorem
Consistent hashing
B-tree
Hash table
JSON
YAML
XML
GraphQL
Representational state transfer
WebSocket
Hypertext Transfer Protocol
HTTPS
Transport Layer Security
Transmission Control Protocol
Domain Name System
Load balancing (computing)
Reverse proxy
Nginx
Apache HTTP Server
Cache (computing)
Content delivery network
Microservices
Serverless computing
Virtual machine
OS-level virtualization
Continuous integration
DevOps
Amazon Web Services
Microsoft Azure
Google Cloud Platform
Git
Linux
Cloud computing
Open-source software
Software as a service
Big data
Edge computing
Internet of things
Cloud storage

# --- Programming languages + software engineering ---
Python (programming language)
Rust (programming language)
JavaScript
TypeScript
Java (programming language)
C (programming language)
C++
C Sharp (programming language)
Go (programming language)
Ruby (programming language)
PHP
Swift (programming language)
Kotlin (programming language)
Scala (programming language)
Haskell
Lisp (programming language)
R (programming language)
MATLAB
Assembly language
HTML
CSS
Node.js
React (software)
Compiler
Interpreter (computing)
Garbage collection (computer science)
Regular expression
Version control
Unit testing
Agile software development
Scrum (software development)
Software design pattern
Object-oriented programming
Functional programming
Recursion (computer science)
Algorithm
Data structure
Big O notation
Time complexity
Sorting algorithm
Binary search
Dynamic programming
Graph theory
Turing machine
Computational complexity theory
P versus NP problem
Operating system
Kernel (operating system)
Unix
MacOS
Microsoft Windows
Android (operating system)
IOS
Web browser
World Wide Web
Internet
Application programming interface
Software bug
Open-source license
GNU General Public License
MIT License
Apache License
Creative Commons

# --- Security + cryptography ---
Cryptography
Public-key cryptography
RSA (cryptosystem)
Hash function
SHA-2
Advanced Encryption Standard
Digital signature
Zero-knowledge proof

# --- Venture / company formation (near-neighbour cluster) ---
Y Combinator
Sequoia Capital
Andreessen Horowitz
Venture capital
Seed money
Initial public offering
Startup company
Angel investor
Private equity
Valuation (finance)
Due diligence
Mergers and acquisitions
Special-purpose acquisition company
Unicorn (finance)
Startup accelerator
Business incubator
Crowdfunding
Kickstarter
Business model
Minimum viable product
Subscription business model
Freemium
Network effect
Economies of scale
Intellectual property
Patent
Trademark
Copyright
Employee stock option

# --- Math / statistics underpinnings ---
Statistics
Probability
Normal distribution
Bayesian inference
Markov chain
Monte Carlo method
Hidden Markov model
Linear algebra
Matrix (mathematics)
Eigenvalues and eigenvectors
Vector space
Calculus
Derivative
Mathematical optimization
Convex optimization
Entropy (information theory)
Information theory
Signal processing
Fourier transform
Fast Fourier transform

# --- Adjacent tech (dilutes the clusters slightly, as a real brain would) ---
Unicode
Blockchain
Cryptocurrency
Bitcoin
Ethereum
Smart contract
Non-fungible token
Quantum computing
Qubit
Quantum entanglement
Semiconductor
Moore's law
Graphics processing unit
Central processing unit
Tensor Processing Unit
Field-programmable gate array
Application-specific integrated circuit
Random-access memory
Solid-state drive
Superconductivity
Nanotechnology
Biotechnology
CRISPR
Genome
Protein folding
Bioinformatics
Neuroscience
Cognitive science
Philosophy of mind
Consciousness
Automation
5G
Wi-Fi
Bluetooth
