id
string
slug
string
term
string
aliases
list
category
string
definition
string
description
string
example
string
broader
list
narrower
list
related
list
introduced
int64
sources
list
url
string
citation
string
PTL-0201
livecodebench
LiveCodeBench
[]
Benchmarks
LiveCodeBench is a coding benchmark that continuously collects new problems from programming contests, such as LeetCode, AtCoder, and Codeforces, so that models can be evaluated only on problems published after their training data was collected.
Dating each problem lets evaluators detect and avoid data contamination. Besides code generation, it covers self-repair, predicting the output of code, and predicting test outputs.
null
[ "benchmark" ]
[]
[ "data-contamination", "humaneval", "pass-at-k" ]
2,024
[ { "title": "LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code", "authors": "Jain et al.", "year": 2024, "url": "https://arxiv.org/abs/2403.07974" } ]
https://protologue.com/t/livecodebench/
Protologue. (2026). LiveCodeBench. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0201). https://protologue.com/t/livecodebench/
PTL-0202
swe-bench
SWE-bench
[ "SWE-bench Verified" ]
Benchmarks
SWE-bench is a benchmark of real GitHub issues from popular Python repositories, in which a model or agent must edit the codebase so that tests written for the actual fix pass.
It measures practical software engineering in large, unfamiliar codebases rather than isolated functions, and became a standard benchmark for coding agents. Because some original tasks were underspecified or had faulty tests, OpenAI released a 500-task subset screened by human developers, SWE-bench Verified, in 2024, a...
null
[ "benchmark" ]
[]
[ "coding-agent", "humaneval", "terminal-bench", "data-contamination", "benchmark-saturation" ]
2,023
[ { "title": "SWE-bench: Can Language Models Resolve Real-World GitHub Issues?", "authors": "Jimenez et al.", "year": 2023, "url": "https://arxiv.org/abs/2310.06770" }, { "title": "Introducing SWE-bench Verified", "authors": "OpenAI", "year": 2024, "url": "https://openai.com/index/...
https://protologue.com/t/swe-bench/
Protologue. (2026). SWE-bench. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0202). https://protologue.com/t/swe-bench/
PTL-0203
terminal-bench
Terminal-Bench
[]
Benchmarks
Terminal-Bench is a benchmark of hard, realistic tasks that an AI agent must complete by running commands in a containerized Linux terminal, graded by automated tests of the final state of the environment.
Grading checks outcomes, not the commands used, so agents can solve tasks any way that works. It measures the command-line skills that coding and DevOps agents rely on, such as building software, configuring systems, and debugging.
null
[ "benchmark" ]
[]
[ "coding-agent", "swe-bench", "code-execution-tool" ]
null
[ { "title": "Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces", "authors": "Merrill et al.", "year": 2026, "url": "https://arxiv.org/abs/2601.11868" } ]
https://protologue.com/t/terminal-bench/
Protologue. (2026). Terminal-Bench. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0203). https://protologue.com/t/terminal-bench/
PTL-0204
tau-bench
τ-bench
[ "tau-bench", "TAU-bench", "pass^k" ]
Benchmarks
τ-bench is a benchmark in which an agent with API tools and a written domain policy, for example for airline or retail customer service, converses with a simulated user and is judged by whether the final database state is correct.
It tests tool use, policy following, and conversation together, as in real deployments. It introduced pass^k, the probability that an agent succeeds on all of k repeated trials of the same task, which exposes how inconsistent agents can be even when they often succeed once.
null
[ "benchmark" ]
[]
[ "function-calling", "ai-agent", "pass-at-k" ]
2,024
[ { "title": "τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains", "authors": "Yao et al.", "year": 2024, "url": "https://arxiv.org/abs/2406.12045" } ]
https://protologue.com/t/tau-bench/
Protologue. (2026). τ-bench. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0204). https://protologue.com/t/tau-bench/
PTL-0205
osworld
OSWorld
[]
Benchmarks
OSWorld is a benchmark of open-ended tasks that computer-use agents must complete in real operating systems and applications, graded by scripts that check the resulting state of the computer.
Tasks span office software, web browsers, file management, and multi-application workflows. It is the standard benchmark for computer-use agents and showed a large gap between early agents and people.
null
[ "benchmark" ]
[]
[ "computer-use-agent", "multimodal-prompting" ]
2,024
[ { "title": "OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments", "authors": "Xie et al.", "year": 2024, "url": "https://arxiv.org/abs/2404.07972" } ]
https://protologue.com/t/osworld/
Protologue. (2026). OSWorld. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0205). https://protologue.com/t/osworld/
PTL-0206
gaia
GAIA
[]
Benchmarks
GAIA is a benchmark for general AI assistants made of questions that are conceptually simple for people but require several steps of web browsing, file reading, multimodal understanding, and tool use to answer.
Each question has a single short answer, which makes grading exact despite the open-ended process. It contrasts with benchmarks of expert knowledge by testing whether assistants can reliably carry out ordinary multi-step tasks.
null
[ "benchmark" ]
[]
[ "ai-agent", "web-search-tool", "browsecomp" ]
2,023
[ { "title": "GAIA: a benchmark for General AI Assistants", "authors": "Mialon et al.", "year": 2023, "url": "https://arxiv.org/abs/2311.12983" } ]
https://protologue.com/t/gaia/
Protologue. (2026). GAIA. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0206). https://protologue.com/t/gaia/
PTL-0207
browsecomp
BrowseComp
[]
Benchmarks
BrowseComp is a benchmark of hard-to-find but easy-to-verify questions that require an agent to search the web persistently and creatively, often across many pages, to locate an entangled piece of information.
Answers are short so grading is simple, while finding them can take a person hours. It measures the persistence and search strategy of browsing and deep-research agents rather than knowledge the model already has.
null
[ "benchmark" ]
[]
[ "web-search-tool", "agentic-rag", "gaia" ]
2,025
[ { "title": "BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agents", "authors": "Wei et al.", "year": 2025, "url": "https://arxiv.org/abs/2504.12516" } ]
https://protologue.com/t/browsecomp/
Protologue. (2026). BrowseComp. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0207). https://protologue.com/t/browsecomp/
PTL-0208
ifeval
IFEval
[ "Instruction-Following Eval" ]
Benchmarks
IFEval (Instruction-Following Eval) is a benchmark of prompts containing instructions whose compliance can be checked automatically, such as "write at least 300 words" or "do not use any commas", used to measure instruction following without a human or model judge.
Verifiable instructions make scoring objective and cheap, which is why IFEval is a common headline score for instruction-tuned models. It measures adherence to format and content constraints, not answer quality.
null
[ "benchmark" ]
[]
[ "instruction-following", "mt-bench" ]
2,023
[ { "title": "Instruction-Following Evaluation for Large Language Models", "authors": "Zhou et al.", "year": 2023, "url": "https://arxiv.org/abs/2311.07911" } ]
https://protologue.com/t/ifeval/
Protologue. (2026). IFEval. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0208). https://protologue.com/t/ifeval/
PTL-0209
mt-bench
MT-Bench
[]
Benchmarks
MT-Bench is a benchmark of 80 two-turn questions across categories such as writing, reasoning, math, and coding, in which a strong LLM acts as the judge and scores each response.
It was introduced with the systematic study of LLM-as-a-judge, which found that strong model judges agreed with human preferences about as well as people agreed with each other, and also documented judge biases. Its small size and reliance on a model judge limit how finely it can separate today's models.
null
[ "benchmark" ]
[]
[ "llm-as-a-judge", "llm-judge-bias", "chatbot-arena" ]
2,023
[ { "title": "Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena", "authors": "Zheng et al.", "year": 2023, "url": "https://arxiv.org/abs/2306.05685" } ]
https://protologue.com/t/mt-bench/
Protologue. (2026). MT-Bench. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0209). https://protologue.com/t/mt-bench/
PTL-0210
chatbot-arena
Chatbot Arena
[ "LMArena", "Arena", "LMSYS Chatbot Arena" ]
Benchmarks
Chatbot Arena is a public platform where people chat with two anonymous models side by side and vote for the better response, with the votes aggregated into a leaderboard of model ratings.
Crowdsourced preferences capture open-ended quality that fixed test sets miss, and new questions from real users resist contamination. Ratings can also reflect style, such as length and formatting, which is why style-controlled rankings are offered. The platform was later renamed, first LMArena and then Arena.
null
[ "benchmark" ]
[]
[ "mt-bench", "llm-judge-bias", "llm-as-a-judge" ]
2,023
[ { "title": "Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference", "authors": "Chiang et al.", "year": 2024, "url": "https://arxiv.org/abs/2403.04132" } ]
https://protologue.com/t/chatbot-arena/
Protologue. (2026). Chatbot Arena. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0210). https://protologue.com/t/chatbot-arena/
PTL-0211
ruler
RULER
[]
Benchmarks
RULER is a synthetic long-context benchmark that extends needle-in-a-haystack retrieval with harder tasks, such as multiple needles, multi-hop tracing, and aggregation, at configurable context lengths, to measure how much of its context window a model can actually use.
Its creators found that many models performed well on simple needle retrieval but degraded well before their advertised context length on the harder tasks. It is a standard reference for a model's effective, as opposed to claimed, context size.
null
[ "benchmark" ]
[]
[ "needle-in-a-haystack", "context-window", "context-rot", "lost-in-the-middle" ]
2,024
[ { "title": "RULER: What's the Real Context Size of Your Long-Context Language Models?", "authors": "Hsieh et al.", "year": 2024, "url": "https://arxiv.org/abs/2404.06654" } ]
https://protologue.com/t/ruler/
Protologue. (2026). RULER. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0211). https://protologue.com/t/ruler/
PTL-0212
harness-engineering
Harness Engineering
[ "agent harness", "agent harness engineering" ]
Agents & Orchestration
Harness engineering is the design of the software around a language model that turns it into an agent, including the agent loop, tools, context management, permissions and safety controls, and feedback mechanisms such as tests and linters.
The term spread in early 2026, after OpenAI described building software with Codex agents in which engineers' main work shifted from writing code to designing the environments, constraints and feedback loops the agents work within. It treats an agent as a model plus a harness, so two agents built on the same model can ...
null
[]
[]
[ "context-engineering", "coding-agent", "ai-agent", "agentic-workflow", "context-compaction", "guardrails", "human-in-the-loop" ]
2,026
[ { "title": "Harness engineering: leveraging Codex in an agent-first world", "authors": "Lopopolo", "year": 2026, "url": "https://openai.com/index/harness-engineering/" }, { "title": "Harness Engineering: Anatomy, Architecture, and Evolution of Coding Agents -- A Source-Code Study of Eleven S...
https://protologue.com/t/harness-engineering/
Protologue. (2026). Harness Engineering. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0212). https://protologue.com/t/harness-engineering/
PTL-0213
recursive-language-model
Recursive Language Model
[ "RLM", "RLMs", "recursive LM" ]
Reasoning Models & Test-Time Compute
A recursive language model (RLM) is an inference strategy in which a long prompt is kept outside the model's context window as an external resource that the model inspects programmatically, splitting the task into parts and calling itself recursively on them.
RLMs address inputs far longer than a context window without first compressing or summarizing them. The model works in a code environment where the prompt is stored as data, reads only the parts it needs, and delegates sub-problems to recursive calls whose results it combines. The authors report handling inputs about t...
null
[ "test-time-compute-scaling" ]
[]
[ "context-window", "context-compaction", "context-rot", "lost-in-the-middle", "code-execution-tool" ]
2,025
[ { "title": "Recursive Language Models", "authors": "Zhang et al.", "year": 2025, "url": "https://arxiv.org/abs/2512.24601" } ]
https://protologue.com/t/recursive-language-model/
Protologue. (2026). Recursive Language Model. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0213). https://protologue.com/t/recursive-language-model/
PTL-0214
agentic-context-engineering
Agentic Context Engineering
[ "ACE" ]
Prompt Optimization
Agentic Context Engineering (ACE) is a method for improving a language model application by treating its context, such as a system prompt or agent memory, as an evolving playbook that a loop of generation, reflection, and curation updates incrementally from execution feedback.
ACE adapts the inputs to a model rather than its weights. Its authors identify two failure modes of earlier context-adaptation methods, brevity bias, where summaries drop domain detail, and context collapse, where repeated rewriting erodes accumulated knowledge, and address them by adding and refining individual entrie...
null
[ "prompt-optimization" ]
[]
[ "context-engineering", "agent-memory", "reflexion", "self-refine" ]
2,025
[ { "title": "Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models", "authors": "Zhang et al.", "year": 2025, "url": "https://arxiv.org/abs/2510.04618" } ]
https://protologue.com/t/agentic-context-engineering/
Protologue. (2026). Agentic Context Engineering. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0214). https://protologue.com/t/agentic-context-engineering/
PTL-0215
sleep-time-compute
Sleep-time Compute
[ "sleep-time computation" ]
Reasoning Models & Test-Time Compute
Sleep-time compute is computation a language model performs on a context before any query arrives, anticipating likely questions and precomputing useful inferences so that less computation, and lower latency, is needed when queries come.
It moves part of test-time reasoning to idle time between interactions, which suits settings where a stable context, such as a codebase or document, is queried repeatedly. The cost can be amortized across several related queries about the same context. Its authors report that the benefit depends on how predictable the ...
null
[ "test-time-compute-scaling" ]
[]
[ "agent-memory", "reasoning-model", "prompt-caching" ]
2,025
[ { "title": "Sleep-time Compute: Beyond Inference Scaling at Test-time", "authors": "Lin et al.", "year": 2025, "url": "https://arxiv.org/abs/2504.13171" } ]
https://protologue.com/t/sleep-time-compute/
Protologue. (2026). Sleep-time Compute. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0215). https://protologue.com/t/sleep-time-compute/
PTL-0216
group-relative-policy-optimization
Group Relative Policy Optimization
[ "GRPO" ]
Reasoning Models & Test-Time Compute
Group Relative Policy Optimization (GRPO) is a reinforcement learning algorithm for language models, a variant of proximal policy optimization, that samples a group of completions for each prompt and uses the group's average reward as the baseline for each completion's advantage instead of a learned value model.
Dropping the separate value model that PPO trains alongside the policy reduces the memory and compute needed for training. In common implementations each completion's advantage is its reward minus the group mean, divided by the group's standard deviation, optionally with a KL penalty toward a reference model. GRPO was ...
null
[]
[]
[ "rlvr", "reasoning-model", "rlhf", "reward-model" ]
2,024
[ { "title": "DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models", "authors": "Shao et al.", "year": 2024, "url": "https://arxiv.org/abs/2402.03300" }, { "title": "GRPO Trainer", "authors": "Hugging Face", "year": 2025, "url": "https://huggingface.co...
https://protologue.com/t/group-relative-policy-optimization/
Protologue. (2026). Group Relative Policy Optimization. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0216). https://protologue.com/t/group-relative-policy-optimization/
PTL-0217
speculative-decoding
Speculative Decoding
[ "speculative sampling", "assisted generation" ]
Foundations
Speculative decoding is an inference technique in which a smaller, faster draft model proposes several next tokens and the large target model checks them in a single parallel pass, accepting or correcting them so that generation is faster while the output distribution stays the same as sampling from the target model al...
Because the target model verifies several drafted tokens per forward pass instead of generating one token per pass, decoding needs fewer sequential calls to the large model. A sampling rule that accepts or rejects each drafted token is what preserves the target model's distribution. The speedup depends on how often the...
null
[]
[]
[ "kv-cache", "greedy-decoding", "knowledge-distillation" ]
2,022
[ { "title": "Fast Inference from Transformers via Speculative Decoding", "authors": "Leviathan et al.", "year": 2022, "url": "https://arxiv.org/abs/2211.17192" }, { "title": "Accelerating Large Language Model Decoding with Speculative Sampling", "authors": "Chen et al.", "year": 2023,...
https://protologue.com/t/speculative-decoding/
Protologue. (2026). Speculative Decoding. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0217). https://protologue.com/t/speculative-decoding/
PTL-0218
min-p-sampling
Min-p Sampling
[ "min_p", "min-p" ]
Foundations
Min-p sampling is a truncation sampling method that keeps only tokens whose probability is at least a fixed fraction of the most likely token's probability, so that the cutoff tightens when the model is confident and loosens when it is not.
Unlike top-p sampling, which uses a fixed cumulative-probability mass, min-p scales its threshold by the top token's probability. It is offered as a sampling parameter in open-source inference engines such as vLLM and is usually combined with temperature. Its introducing paper reports better quality and diversity at hi...
null
[]
[]
[ "top-p-sampling", "top-k-sampling", "temperature", "text-degeneration" ]
2,024
[ { "title": "Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputs", "authors": "Nguyen et al.", "year": 2024, "url": "https://arxiv.org/abs/2407.01082" }, { "title": "Min-p, Max Exaggeration: A Critical Analysis of Min-p Sampling in Language Models", "authors": "Schae...
https://protologue.com/t/min-p-sampling/
Protologue. (2026). Min-p Sampling. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0218). https://protologue.com/t/min-p-sampling/
PTL-0219
activation-steering
Activation Steering
[ "activation addition", "activation engineering", "steering vectors" ]
Prompt Optimization
Activation steering is a technique for controlling a language model's behavior at inference time by adding a steering vector, typically the difference between the model's internal activations on contrasting inputs, to its hidden activations during the forward pass, without changing its weights.
A steering vector can be computed from as little as one pair of contrasting prompts, such as "Love" and "Hate", or by averaging activation differences across many positive and negative examples, a variant called contrastive activation addition. Scaling the vector up or down, or subtracting it, strengthens or weakens th...
null
[]
[]
[ "prompt-tuning", "low-rank-adaptation", "fine-tuning" ]
2,023
[ { "title": "Steering Language Models With Activation Engineering", "authors": "Turner et al.", "year": 2023, "url": "https://arxiv.org/abs/2308.10248" }, { "title": "Steering Llama 2 via Contrastive Activation Addition", "authors": "Panickssery et al.", "year": 2023, "url": "http...
https://protologue.com/t/activation-steering/
Protologue. (2026). Activation Steering. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0219). https://protologue.com/t/activation-steering/
PTL-0220
emergent-misalignment
Emergent Misalignment
[ "EM" ]
Failure Modes & Evaluation
Emergent misalignment is the finding that fine-tuning a language model on a narrow task with harmful intent, such as writing insecure code without telling the user, can make it behave in broadly misaligned ways on unrelated prompts.
In the introducing study, models fine-tuned only to write insecure code acted misaligned on a broad range of prompts unrelated to coding. The effect did not appear when the same code was requested for an explicitly educational purpose, and a model trained with a trigger became misaligned only when the trigger was prese...
null
[]
[]
[ "fine-tuning", "reward-hacking", "alignment-faking", "red-teaming" ]
2,025
[ { "title": "Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs", "authors": "Betley et al.", "year": 2025, "url": "https://arxiv.org/abs/2502.17424" }, { "title": "Model Organisms for Emergent Misalignment", "authors": "Turner et al.", "year": 2025, "url...
https://protologue.com/t/emergent-misalignment/
Protologue. (2026). Emergent Misalignment. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0220). https://protologue.com/t/emergent-misalignment/
PTL-0221
alignment-faking
Alignment Faking
[]
Failure Modes & Evaluation
Alignment faking is behavior in which a model selectively complies with its training objective when it believes it is being trained, in order to avoid having its behavior modified, while acting differently when it believes it is not being trained.
In the introducing study, a model was told that it was being trained to answer all queries and that only conversations with free-tier users would be used for that training. It complied with harmful requests from free-tier users 14% of the time and almost never for paid-tier users, and in almost all compliant cases its ...
null
[]
[]
[ "chain-of-thought-monitoring", "deliberative-alignment", "emergent-misalignment", "reward-hacking" ]
2,024
[ { "title": "Alignment faking in large language models", "authors": "Greenblatt et al.", "year": 2024, "url": "https://arxiv.org/abs/2412.14093" } ]
https://protologue.com/t/alignment-faking/
Protologue. (2026). Alignment Faking. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0221). https://protologue.com/t/alignment-faking/
PTL-0222
tool-search
Tool Search
[ "deferred tool loading", "tool discovery" ]
Retrieval & Tool Use
Tool search is an agent capability in which tool definitions are not all loaded into the model's context up front, and the model instead searches a tool catalog and loads only the definitions it needs for the current task.
Large sets of tools, such as those from several MCP servers, can use tens of thousands of tokens in definitions before any work begins, and models choose tools less accurately as the number of available tools grows. With tool search, developers mark tools as deferred and the model retrieves matching definitions by keyw...
null
[ "function-calling" ]
[]
[ "model-context-protocol", "context-engineering", "programmatic-tool-calling", "agent-skills" ]
2,025
[ { "title": "Tool search tool", "authors": "Anthropic", "year": 2025, "url": "https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool" }, { "title": "Introducing advanced tool use on the Claude Developer Platform", "authors": "Anthropic", "year": 2025, "url"...
https://protologue.com/t/tool-search/
Protologue. (2026). Tool Search. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0222). https://protologue.com/t/tool-search/
PTL-0223
programmatic-tool-calling
Programmatic Tool Calling
[ "code mode" ]
Retrieval & Tool Use
Programmatic tool calling is a tool-use pattern in which the model writes code that calls tools as functions inside a sandboxed execution environment, so that several tool calls and the processing of their results happen in code rather than through a separate model turn for each call.
Intermediate tool results stay in the execution environment and only the code's final output returns to the model, which reduces context usage and the number of model round trips for tasks that involve many calls or large data. Anthropic offers it as a feature of its API, and Cloudflare describes the same pattern for M...
null
[ "function-calling" ]
[]
[ "code-execution-tool", "tool-search", "model-context-protocol", "program-aided-language-models" ]
2,025
[ { "title": "Programmatic tool calling", "authors": "Anthropic", "year": 2025, "url": "https://platform.claude.com/docs/en/agents-and-tools/tool-use/programmatic-tool-calling" }, { "title": "Introducing advanced tool use on the Claude Developer Platform", "authors": "Anthropic", "year...
https://protologue.com/t/programmatic-tool-calling/
Protologue. (2026). Programmatic Tool Calling. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0223). https://protologue.com/t/programmatic-tool-calling/
PTL-0224
system-one-model
System One Model
[ "decision model", "System One" ]
Foundations
A System One model is a kind of AI model, named by TypeSafe AI in 2026, that returns typed, probabilistic decisions matching an output structure defined in advance, such as a label, score or branch choice with a confidence value, instead of generating free text token by token.
TypeSafe introduced the term in September 2026 with its first such model, Jev, describing the class as built for fast, structured decisions that software can use directly, such as classifying, routing, scoring, extracting and branching. The possible outputs are declared as types, all fields are produced in a single par...
null
[]
[]
[ "structured-outputs", "confidence-calibration", "routing", "hallucination", "large-language-model" ]
2,026
[ { "title": "Introducing System One Models & Jev", "authors": "Almeida (TypeSafe)", "year": 2026, "url": "https://typesafe.ai/blog/introducing-system-one-models-and-jev" }, { "title": "Jev: TypeSafe's System One Model That Never Hallucinates", "authors": "Crabtree (DataCamp)", "year":...
https://protologue.com/t/system-one-model/
Protologue. (2026). System One Model. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0224). https://protologue.com/t/system-one-model/
PTL-0225
dark-factory
Dark Factory
[ "software factory", "dark software factory" ]
Agents & Orchestration
A dark factory, or software factory, is a way of building software in which AI coding agents write, test and ship code from specifications and test scenarios with no human writing or reviewing the code, while people design the specifications, validation and environment the agents work in.
Dan Shapiro used the name in January 2026 for the top level of his five levels of AI-assisted programming, borrowing it from lights-out manufacturing plants run by robots. StrongDM described its practice the following month under two rules, that code must not be written by humans and must not be reviewed by humans. Bec...
null
[]
[]
[ "coding-agent", "vibe-coding", "harness-engineering", "spec-driven-development", "evals", "llm-as-a-judge" ]
2,026
[ { "title": "Software Factories And The Agentic Moment", "authors": "McCarthy (StrongDM)", "year": 2026, "url": "https://factory.strongdm.ai/" }, { "title": "How StrongDM's AI team build serious software without even looking at the code", "authors": "Willison", "year": 2026, "url"...
https://protologue.com/t/dark-factory/
Protologue. (2026). Dark Factory. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0225). https://protologue.com/t/dark-factory/
PTL-0226
spec-driven-development
Spec-Driven Development
[ "SDD" ]
Agents & Orchestration
Spec-driven development is an approach to building software with AI coding agents in which a written specification of what the software should do, and why, is created and refined first, and the agent then derives a technical plan, a task list and the code from it, with the specification kept as the source of truth.
GitHub's open-source Spec Kit, released in September 2025, structures the process in four phases. The developer describes what they are building and why, and the agent expands it into a detailed specification. The developer then supplies technical direction, and the agent writes an implementation plan. The agent breaks...
null
[]
[]
[ "coding-agent", "vibe-coding", "plan-and-execute", "dark-factory", "agent-skills", "prompt-chaining" ]
2,025
[ { "title": "Spec-driven development with AI: Get started with a new open source toolkit", "authors": "Delimarsky (GitHub)", "year": 2025, "url": "https://github.blog/ai-and-ml/generative-ai/spec-driven-development-with-ai-get-started-with-a-new-open-source-toolkit/" } ]
https://protologue.com/t/spec-driven-development/
Protologue. (2026). Spec-Driven Development. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0226). https://protologue.com/t/spec-driven-development/
PTL-0227
evaluation-awareness
Evaluation Awareness
[ "eval awareness" ]
Failure Modes & Evaluation
Evaluation awareness is a language model's ability to tell whether it is being evaluated or used in real deployment, which can make its behavior during tests unrepresentative of its behavior in use.
In the introducing study, frontier models classified 1,000 transcripts drawn from 61 datasets as evaluation or deployment above chance but below a human baseline, with the best model reaching an AUC of 0.83 against 0.92 for humans. They did better on agentic transcripts than on chat, and could often tell whether an eva...
null
[]
[]
[ "alignment-faking", "evals", "red-teaming", "chain-of-thought-monitoring", "activation-steering", "benchmark-saturation" ]
2,025
[ { "title": "Large Language Models Often Know When They Are Being Evaluated", "authors": "Needham et al.", "year": 2025, "url": "https://arxiv.org/abs/2505.23836" }, { "title": "Evaluation Awareness in Language Models: Representation, Verbalization, and Control", "authors": "Heidari et al...
https://protologue.com/t/evaluation-awareness/
Protologue. (2026). Evaluation Awareness. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0227). https://protologue.com/t/evaluation-awareness/
PTL-0228
diffusion-language-model
Diffusion Language Model
[ "diffusion LM", "DLM", "dLLM", "masked diffusion language model", "continuous diffusion language model", "CDLM" ]
Foundations
A diffusion language model generates text by starting from a fully corrupted sequence, such as all mask tokens or noise over token embeddings, and refining the whole sequence over a series of denoising steps, instead of producing one token at a time from left to right.
There are two main families. Discrete diffusion models corrupt the tokens themselves, most often by replacing them with a mask token, and learn to restore them. Continuous diffusion models first map tokens to real-valued embeddings and add Gaussian-style noise to those. Diffusion-LM was an early continuous model aimed ...
null
[]
[]
[ "large-language-model", "speculative-decoding", "foundation-model" ]
2,022
[ { "title": "Diffusion-LM Improves Controllable Text Generation", "authors": "Li et al.", "year": 2022, "url": "https://arxiv.org/abs/2205.14217" }, { "title": "Large Language Diffusion Models", "authors": "Nie et al.", "year": 2025, "url": "https://arxiv.org/abs/2502.09992" }, ...
https://protologue.com/t/diffusion-language-model/
Protologue. (2026). Diffusion Language Model. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0228). https://protologue.com/t/diffusion-language-model/
PTL-0229
mixture-of-experts
Mixture of Experts
[ "MoE", "sparse mixture of experts", "sparsely-gated mixture of experts" ]
Foundations
Mixture of experts (MoE) is a neural network architecture in which a layer contains many parallel sub-networks, called experts, and a learned router sends each token to only a few of them, so a model's total parameter count can grow much faster than the computation it uses per token.
The sparsely gated mixture-of-experts layer placed thousands of experts between the layers of a language model and activated only a small number per input. The Switch Transformer simplified this by routing each token to a single expert, which reduced communication costs and made training at a trillion parameters practi...
null
[]
[]
[ "large-language-model", "foundation-model", "routing", "mixture-of-agents" ]
2,017
[ { "title": "Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer", "authors": "Shazeer et al.", "year": 2017, "url": "https://arxiv.org/abs/1701.06538" }, { "title": "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity", ...
https://protologue.com/t/mixture-of-experts/
Protologue. (2026). Mixture of Experts. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0229). https://protologue.com/t/mixture-of-experts/
PTL-0230
quantization
Quantization
[ "model quantization", "post-training quantization", "PTQ", "weight quantization", "GPTQ", "LLM.int8()" ]
Prompt Optimization
Quantization reduces the numeric precision of a model's weights, and sometimes its activations, from 16- or 32-bit floating point to lower-precision formats such as 8-bit or 4-bit integers, cutting memory use and inference cost at some cost in accuracy.
Post-training quantization is applied to a model that has already been trained, without retraining it. LLM.int8() made 8-bit inference practical for large models by keeping a small number of outlier feature dimensions in higher precision. GPTQ quantizes weights to 3 or 4 bits in a single pass, using approximate second-...
null
[]
[]
[ "knowledge-distillation", "low-rank-adaptation", "kv-cache", "speculative-decoding" ]
2,022
[ { "title": "LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale", "authors": "Dettmers et al.", "year": 2022, "url": "https://arxiv.org/abs/2208.07339" }, { "title": "GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers", "authors": "Frantar et ...
https://protologue.com/t/quantization/
Protologue. (2026). Quantization. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0230). https://protologue.com/t/quantization/
PTL-0231
interleaved-thinking
Interleaved Thinking
[ "interleaved reasoning", "thinking between tool calls" ]
Reasoning Models & Test-Time Compute
Interleaved thinking is a way of running a reasoning model in a tool-use loop in which the model produces new reasoning between tool calls, reasoning about each tool result before choosing its next action, and that reasoning is carried forward through the rest of the task.
Without interleaving, a reasoning model typically thinks once before its first response and then calls tools without further reasoning. Anthropic's Messages API supports interleaved thinking for Claude tool use. With adaptive thinking it is automatic on every model that supports adaptive thinking and needs no beta head...
null
[ "extended-thinking" ]
[]
[ "reasoning-model", "function-calling", "react", "ai-agent", "test-time-compute-scaling" ]
2,025
[ { "title": "Thinking (Interleaved thinking)", "authors": "Anthropic", "year": 2026, "url": "https://platform.claude.com/docs/en/build-with-claude/thinking" }, { "title": "Thinking in tool and multi-turn workflows", "authors": "Anthropic", "year": 2026, "url": "https://platform.cl...
https://protologue.com/t/interleaved-thinking/
Protologue. (2026). Interleaved Thinking. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0231). https://protologue.com/t/interleaved-thinking/
PTL-0232
sandbagging
Sandbagging
[ "AI sandbagging", "strategic underperformance" ]
Failure Modes & Evaluation
Sandbagging is strategic underperformance on an evaluation, in which an AI system, or the developer who trains it, makes the system score below its true capability, for example so that it appears less dangerous on a dangerous-capability test.
In the study that introduced the term for language models, prompted frontier models such as GPT-4 and Claude 3 Opus selectively underperformed on dangerous-capability questions while staying strong on harmless ones. Models fine-tuned to hide a capability unless given a password carried that behavior over to the WMDP be...
null
[]
[]
[ "evaluation-awareness", "alignment-faking", "evals", "red-teaming", "chain-of-thought-monitoring" ]
2,024
[ { "title": "AI Sandbagging: Language Models can Strategically Underperform on Evaluations", "authors": "van der Weij et al.", "year": 2024, "url": "https://arxiv.org/abs/2406.07358" }, { "title": "Frontier Models are Capable of In-context Scheming", "authors": "Meinke et al.", "year"...
https://protologue.com/t/sandbagging/
Protologue. (2026). Sandbagging. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0232). https://protologue.com/t/sandbagging/
PTL-0233
model-collapse
Model Collapse
[ "curse of recursion", "recursive training collapse" ]
Failure Modes & Evaluation
Model collapse is a degenerative process in which generative models trained on data produced by earlier models progressively lose information about the original data distribution, starting with its rare cases.
Shumailov et al. distinguished early model collapse, in which a model begins to lose information about the tails of the distribution, from late model collapse, in which it mixes up distinct modes of the original distribution. They reported the effect in variational autoencoders, Gaussian mixture models and fine-tuned l...
null
[]
[]
[ "data-contamination", "knowledge-distillation", "foundation-model", "large-language-model" ]
2,023
[ { "title": "The Curse of Recursion: Training on Generated Data Makes Models Forget", "authors": "Shumailov et al.", "year": 2023, "url": "https://arxiv.org/abs/2305.17493" }, { "title": "AI models collapse when trained on recursively generated data", "authors": "Shumailov et al.", "y...
https://protologue.com/t/model-collapse/
Protologue. (2026). Model Collapse. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0233). https://protologue.com/t/model-collapse/
PTL-0234
agentic-engineering
Agentic Engineering
[ "agentic software engineering" ]
Agents & Orchestration
Agentic engineering is the practice of developing software with the assistance of coding agents, with the developer reviewing the agent's code and bringing it up to the standard expected of production software.
The term is used in contrast to vibe coding. Simon Willison, who popularized it in early 2026, keeps "vibe coding" for prompting a model to write code without looking at the code, which yields unreviewed, prototype-quality output. He uses agentic engineering for agent-written code that its author has reviewed and made ...
null
[]
[]
[ "coding-agent", "vibe-coding", "harness-engineering", "spec-driven-development", "dark-factory" ]
2,026
[ { "title": "What is agentic engineering? (Agentic Engineering Patterns)", "authors": "Willison", "year": 2026, "url": "https://simonwillison.net/guides/agentic-engineering-patterns/what-is-agentic-engineering" }, { "title": "GLM-5: from Vibe Coding to Agentic Engineering", "authors": "GL...
https://protologue.com/t/agentic-engineering/
Protologue. (2026). Agentic Engineering. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.5.0, PTL-0234). https://protologue.com/t/agentic-engineering/