franklinpolymathic commited on
Commit
e503b25
·
verified ·
1 Parent(s): 6102a65

Protologue v1.5.0 from GitHub release

Browse files
Files changed (2) hide show
  1. data/categories.jsonl +3 -3
  2. data/terms.jsonl +0 -0
data/categories.jsonl CHANGED
@@ -3,9 +3,9 @@
3
  {"id": "reasoning", "name": "Reasoning Elicitation", "description": "Techniques that get a model to produce intermediate reasoning, decompose problems, or explore multiple solution paths before answering.", "url": "https://protologue.com/c/reasoning/", "terms": ["Chain-of-Thought Prompting", "Zero-shot Chain-of-Thought", "Scratchpad", "Automatic Chain-of-Thought", "Self-Consistency", "Universal Self-Consistency", "Least-to-Most Prompting", "Plan-and-Solve Prompting", "Step-Back Prompting", "Tree of Thoughts", "Graph of Thoughts", "Skeleton-of-Thought", "Analogical Prompting", "Self-Discover", "Contrastive Chain-of-Thought", "Thread of Thought", "System 2 Attention", "Rephrase and Respond", "Self-Ask", "Maieutic Prompting", "Generated Knowledge Prompting", "Program-Aided Language Models", "Program of Thoughts", "Decomposed Prompting", "Complexity-Based Prompting", "Chain of Draft"]}
4
  {"id": "verification", "name": "Self-Critique & Verification", "description": "Techniques in which a model, or a set of models, checks, critiques, votes on, or revises outputs.", "url": "https://protologue.com/c/verification/", "terms": ["Self-Refine", "Reflexion", "Chain-of-Verification", "LLM-as-a-Judge", "Multi-Agent Debate", "Mixture-of-Agents", "Process Reward Model", "Best-of-N Sampling", "Self-Correction", "Self-Verification", "CRITIC", "Chain of Density", "Confidence Calibration", "SelfCheckGPT"]}
5
  {"id": "retrieval-tools", "name": "Retrieval & Tool Use", "description": "Grounding generation in external information and letting models call functions, search, and other tools.", "url": "https://protologue.com/c/retrieval-tools/", "terms": ["Retrieval-Augmented Generation", "Hypothetical Document Embeddings", "Self-RAG", "ReAct", "Toolformer", "Function Calling", "Model Context Protocol", "Structured Outputs", "Embedding", "Vector Database", "Chunking", "Dense Retrieval", "BM25", "Hybrid Search", "Reranking", "ColBERT", "Query Rewriting", "Contextual Retrieval", "GraphRAG", "Agentic RAG", "Corrective RAG", "Active Retrieval", "Web Search Tool", "Code Execution Tool", "Citation Generation", "Tool Search", "Programmatic Tool Calling"]}
6
- {"id": "agents", "name": "Agents & Orchestration", "description": "Patterns for composing multiple model calls, tools, and memory into workflows and autonomous agents.", "url": "https://protologue.com/c/agents/", "terms": ["AI Agent", "Agentic Workflow", "Prompt Chaining", "Routing", "Parallelization", "Orchestrator-Workers", "Evaluator-Optimizer", "Meta-Prompting", "Agent Memory", "Context Engineering", "Multi-Agent System", "Plan-and-Execute", "Handoff", "Human-in-the-Loop", "Computer-Use Agent", "Coding Agent", "Vibe Coding", "Agent2Agent Protocol", "Agent Skills", "Context Compaction", "Harness Engineering", "Dark Factory", "Spec-Driven Development"]}
7
  {"id": "optimization", "name": "Prompt Optimization", "description": "Automatic and learned methods that search for, compress, or train better prompts.", "url": "https://protologue.com/c/optimization/", "terms": ["Automatic Prompt Engineer", "Optimization by Prompting", "DSPy", "Prompt Tuning", "Prefix Tuning", "Low-Rank Adaptation", "Directional Stimulus Prompting", "Prompt Compression", "Prompt Caching", "Emotion Prompting", "Prompt Optimization", "ProTeGi", "TextGrad", "Promptbreeder", "EvoPrompt", "Model Cascade", "Agentic Context Engineering", "Activation Steering", "Quantization"]}
8
- {"id": "test-time", "name": "Reasoning Models & Test-Time Compute", "description": "Models trained to reason at length, and methods that improve answers by spending more computation at inference time.", "url": "https://protologue.com/c/test-time/", "terms": ["Reasoning Model", "Test-Time Compute Scaling", "Extended Thinking", "Self-Taught Reasoner", "Reinforcement Learning with Verifiable Rewards", "Budget Forcing", "Chain-of-Thought Monitoring", "Recursive Language Model", "Sleep-time Compute", "Group Relative Policy Optimization"]}
9
  {"id": "security", "name": "Security & Adversarial Prompting", "description": "Attacks that subvert a model's instructions and the defenses designed to resist them.", "url": "https://protologue.com/c/security/", "terms": ["Prompt Injection", "Indirect Prompt Injection", "Prompt Leaking", "Jailbreak", "Adversarial Suffix", "Many-shot Jailbreaking", "Instruction Hierarchy", "Spotlighting", "Constitutional AI", "Red Teaming", "Guardrails", "Dual LLM Pattern", "Tool Poisoning", "Lethal Trifecta", "Crescendo Attack", "PAIR", "Excessive Agency", "Deliberative Alignment", "Constitutional Classifiers"]}
10
- {"id": "failure-modes", "name": "Failure Modes & Evaluation", "description": "Systematic ways prompted models go wrong, and the evaluations used to measure them.", "url": "https://protologue.com/c/failure-modes/", "terms": ["Hallucination", "Sycophancy", "Lost in the Middle", "Needle in a Haystack", "Prompt Sensitivity", "Unfaithful Chain-of-Thought", "Context Rot", "Overthinking", "Reversal Curse", "Reward Hacking", "Irrelevant Context Distraction", "Multi-Turn Degradation", "Text Degeneration", "Over-Refusal", "LLM Judge Bias", "Data Contamination", "Evals", "pass@k", "RAG Evaluation", "Emergent Misalignment", "Alignment Faking", "Evaluation Awareness"]}
11
  {"id": "benchmarks", "name": "Benchmarks", "description": "Standard test sets used to measure and compare language models, what each one measures, how it is scored, the prompting setup scores are reported under, and its known limitations.", "url": "https://protologue.com/c/benchmarks/", "terms": ["Benchmark", "Benchmark Saturation", "MMLU", "MMLU-Pro", "GPQA", "Humanity's Last Exam", "BIG-Bench Hard", "ARC-AGI", "SimpleQA", "TruthfulQA", "HellaSwag", "MMMU", "GSM8K", "MATH", "AIME", "HumanEval", "LiveCodeBench", "SWE-bench", "Terminal-Bench", "τ-bench", "OSWorld", "GAIA", "BrowseComp", "IFEval", "MT-Bench", "Chatbot Arena", "RULER"]}
 
3
  {"id": "reasoning", "name": "Reasoning Elicitation", "description": "Techniques that get a model to produce intermediate reasoning, decompose problems, or explore multiple solution paths before answering.", "url": "https://protologue.com/c/reasoning/", "terms": ["Chain-of-Thought Prompting", "Zero-shot Chain-of-Thought", "Scratchpad", "Automatic Chain-of-Thought", "Self-Consistency", "Universal Self-Consistency", "Least-to-Most Prompting", "Plan-and-Solve Prompting", "Step-Back Prompting", "Tree of Thoughts", "Graph of Thoughts", "Skeleton-of-Thought", "Analogical Prompting", "Self-Discover", "Contrastive Chain-of-Thought", "Thread of Thought", "System 2 Attention", "Rephrase and Respond", "Self-Ask", "Maieutic Prompting", "Generated Knowledge Prompting", "Program-Aided Language Models", "Program of Thoughts", "Decomposed Prompting", "Complexity-Based Prompting", "Chain of Draft"]}
4
  {"id": "verification", "name": "Self-Critique & Verification", "description": "Techniques in which a model, or a set of models, checks, critiques, votes on, or revises outputs.", "url": "https://protologue.com/c/verification/", "terms": ["Self-Refine", "Reflexion", "Chain-of-Verification", "LLM-as-a-Judge", "Multi-Agent Debate", "Mixture-of-Agents", "Process Reward Model", "Best-of-N Sampling", "Self-Correction", "Self-Verification", "CRITIC", "Chain of Density", "Confidence Calibration", "SelfCheckGPT"]}
5
  {"id": "retrieval-tools", "name": "Retrieval & Tool Use", "description": "Grounding generation in external information and letting models call functions, search, and other tools.", "url": "https://protologue.com/c/retrieval-tools/", "terms": ["Retrieval-Augmented Generation", "Hypothetical Document Embeddings", "Self-RAG", "ReAct", "Toolformer", "Function Calling", "Model Context Protocol", "Structured Outputs", "Embedding", "Vector Database", "Chunking", "Dense Retrieval", "BM25", "Hybrid Search", "Reranking", "ColBERT", "Query Rewriting", "Contextual Retrieval", "GraphRAG", "Agentic RAG", "Corrective RAG", "Active Retrieval", "Web Search Tool", "Code Execution Tool", "Citation Generation", "Tool Search", "Programmatic Tool Calling"]}
6
+ {"id": "agents", "name": "Agents & Orchestration", "description": "Patterns for composing multiple model calls, tools, and memory into workflows and autonomous agents.", "url": "https://protologue.com/c/agents/", "terms": ["AI Agent", "Agentic Workflow", "Prompt Chaining", "Routing", "Parallelization", "Orchestrator-Workers", "Evaluator-Optimizer", "Meta-Prompting", "Agent Memory", "Context Engineering", "Multi-Agent System", "Plan-and-Execute", "Handoff", "Human-in-the-Loop", "Computer-Use Agent", "Coding Agent", "Vibe Coding", "Agent2Agent Protocol", "Agent Skills", "Context Compaction", "Harness Engineering", "Dark Factory", "Spec-Driven Development", "Agentic Engineering"]}
7
  {"id": "optimization", "name": "Prompt Optimization", "description": "Automatic and learned methods that search for, compress, or train better prompts.", "url": "https://protologue.com/c/optimization/", "terms": ["Automatic Prompt Engineer", "Optimization by Prompting", "DSPy", "Prompt Tuning", "Prefix Tuning", "Low-Rank Adaptation", "Directional Stimulus Prompting", "Prompt Compression", "Prompt Caching", "Emotion Prompting", "Prompt Optimization", "ProTeGi", "TextGrad", "Promptbreeder", "EvoPrompt", "Model Cascade", "Agentic Context Engineering", "Activation Steering", "Quantization"]}
8
+ {"id": "test-time", "name": "Reasoning Models & Test-Time Compute", "description": "Models trained to reason at length, and methods that improve answers by spending more computation at inference time.", "url": "https://protologue.com/c/test-time/", "terms": ["Reasoning Model", "Test-Time Compute Scaling", "Extended Thinking", "Self-Taught Reasoner", "Reinforcement Learning with Verifiable Rewards", "Budget Forcing", "Chain-of-Thought Monitoring", "Recursive Language Model", "Sleep-time Compute", "Group Relative Policy Optimization", "Interleaved Thinking"]}
9
  {"id": "security", "name": "Security & Adversarial Prompting", "description": "Attacks that subvert a model's instructions and the defenses designed to resist them.", "url": "https://protologue.com/c/security/", "terms": ["Prompt Injection", "Indirect Prompt Injection", "Prompt Leaking", "Jailbreak", "Adversarial Suffix", "Many-shot Jailbreaking", "Instruction Hierarchy", "Spotlighting", "Constitutional AI", "Red Teaming", "Guardrails", "Dual LLM Pattern", "Tool Poisoning", "Lethal Trifecta", "Crescendo Attack", "PAIR", "Excessive Agency", "Deliberative Alignment", "Constitutional Classifiers"]}
10
+ {"id": "failure-modes", "name": "Failure Modes & Evaluation", "description": "Systematic ways prompted models go wrong, and the evaluations used to measure them.", "url": "https://protologue.com/c/failure-modes/", "terms": ["Hallucination", "Sycophancy", "Lost in the Middle", "Needle in a Haystack", "Prompt Sensitivity", "Unfaithful Chain-of-Thought", "Context Rot", "Overthinking", "Reversal Curse", "Reward Hacking", "Irrelevant Context Distraction", "Multi-Turn Degradation", "Text Degeneration", "Over-Refusal", "LLM Judge Bias", "Data Contamination", "Evals", "pass@k", "RAG Evaluation", "Emergent Misalignment", "Alignment Faking", "Evaluation Awareness", "Sandbagging", "Model Collapse"]}
11
  {"id": "benchmarks", "name": "Benchmarks", "description": "Standard test sets used to measure and compare language models, what each one measures, how it is scored, the prompting setup scores are reported under, and its known limitations.", "url": "https://protologue.com/c/benchmarks/", "terms": ["Benchmark", "Benchmark Saturation", "MMLU", "MMLU-Pro", "GPQA", "Humanity's Last Exam", "BIG-Bench Hard", "ARC-AGI", "SimpleQA", "TruthfulQA", "HellaSwag", "MMMU", "GSM8K", "MATH", "AIME", "HumanEval", "LiveCodeBench", "SWE-bench", "Terminal-Bench", "τ-bench", "OSWorld", "GAIA", "BrowseComp", "IFEval", "MT-Bench", "Chatbot Arena", "RULER"]}
data/terms.jsonl CHANGED
The diff for this file is too large to render. See raw diff