Skip to content

LlmWikis knowledge page

High-Priority Article Backlog

The audit recommends starting with the articles that anchor reader understanding of large language models. This backlog turns the first 20 candidates into a concrete editorial queue.

First 20-page batch

Page Issue First edits Source targets
Transformer architecture Stub, unsourced Definition, self-attention, multi-head attention, diagram, limitations. Original Transformer paper, LLM surveys.
Self-attention mechanism Thin concept page Query/key/value explanation, intuition, multi-head benefit, examples. Transformer paper, attention explainers.
Tokenization Stub BPE, WordPiece, SentencePiece, vocabulary tradeoffs, examples. Tokenizer docs and model papers.
Pre-training Incomplete Language modeling objective, corpus scale, contrast with supervised training. GPT, BERT, and survey sources.
Fine-tuning Stub Full fine-tuning, adapters, prompt tuning, task examples. BERT, PEFT, provider docs.
RLHF Outdated Preference data, reward model, PPO or alternatives, limitations. OpenAI alignment sources, surveys.
GPT-4 Outdated info Release context, public capabilities, limitations, official claims only. OpenAI technical report and docs.
LLaMA Stub Model family, release context, open-weight posture, variants. Meta papers and official release notes.
BERT Brief Encoder-only architecture, masked language modeling, historical impact. BERT paper and NLP references.
Chatbots Thin overview LLM chat architecture, turn-taking, memory limits, safety controls. Conversational AI surveys and official docs.
Summarization Stub Extractive vs abstractive, prompt-based workflows, evaluation caveats. Summarization papers and benchmark docs.
Code generation Stub Assistant behavior, examples, risk, evaluation, licensing concerns. Codex/Copilot research and docs.
Hugging Face Transformers Thin Library purpose, hub, tokenizers, pipeline example, limitations. Official Hugging Face docs.
LangChain and RAG Thin Chains, retrieval, agents, vector stores, when to avoid complexity. LangChain docs and RAG papers.
Hallucination Stub Definition, causes, mitigations, evaluation, examples. Reliability papers and safety references.
Bias in LLMs Stub Bias sources, measurement, mitigation, limitations. Bias benchmarks and research papers.
AI alignment Stub Alignment definition, RLHF relationship, goals, limits. Alignment literature and official policies.
Privacy and data Stub Training data, memorization, PII risks, privacy attacks, compliance. Memorization and privacy research.
Evaluation metrics Technical stub Perplexity, BLEU, ROUGE, human eval, benchmark caveats. Standard NLP evaluation papers.
Prompt engineering Stub Definition, examples, few-shot prompting, failure modes. GPT-3 paper and provider guides.

Execution order

  1. Week 1: classify all 20 pages and attach source leads before drafting.
  2. Weeks 2-4: expand the seven high-impact foundation/training/model pages.
  3. Weeks 5-7: expand application, tooling, safety, and evaluation pages.
  4. Weeks 8-9: peer review, copy edit, add internal links, and update quality classes.