From prompt-repeater
Optimizes LLM prompts via Google Research-backed repetition for non-reasoning tasks like multiple-choice questions, list navigation, and fact retrieval to boost performance without latency.
npx claudepluginhub danielraffel/generous-corp-marketplace --plugin prompt-repeaterThis skill uses the workspace's default tool permissions.
Prompt repetition is a simple yet effective technique discovered by Google Research that improves LLM performance on non-reasoning tasks by enabling each prompt token to attend to every other prompt token. The technique transforms `<QUERY>` into `<QUERY><QUERY>`, allowing bidirectional attention in causal language models.
Optimizes prompts for any LLM using RSCIT framework, chain-of-thought, few-shot examples, and structured output patterns to improve quality, reduce hallucinations, and cut token usage.
Provides workflows to write, debug, and optimize LLM prompts using few-shot examples, chain-of-thought structuring, system prompts, and templates. Activates for prompt improvement requests.
Optimizes LLM prompts using constitutional AI, chain-of-thought reasoning, and model-specific techniques. Provides guidance, best practices, and checklists for prompt engineering workflows.
Share bugs, ideas, or general feedback.
Prompt repetition is a simple yet effective technique discovered by Google Research that improves LLM performance on non-reasoning tasks by enabling each prompt token to attend to every other prompt token. The technique transforms <QUERY> into <QUERY><QUERY>, allowing bidirectional attention in causal language models.
Key benefits:
Apply prompt repetition to tasks that don't require complex multi-step reasoning:
Multiple Choice Questions:
List Navigation:
Simple Queries:
Pattern Recognition:
Avoid emphasizing repetition for tasks requiring complex reasoning:
Note: Prompt repetition is neutral to slightly positive even with reasoning enabled (5 wins, 1 loss, 22 neutral), so it's safe to use but provides minimal benefit.
When user submits a prompt, evaluate if it's a non-reasoning task:
Non-reasoning indicators:
Reasoning indicators:
For non-reasoning tasks, suggest using one of the available commands:
Simple Repetition (/repeat-last):
<QUERY><QUERY>Verbose Repetition (/repeat-verbose):
<QUERY> Let me repeat that: <QUERY>Triple Repetition (/repeat-3x):
<QUERY> Let me repeat that: <QUERY> Let me repeat that one more time: <QUERY>User asks: "Which of these options is correct? A) ... B) ... C) ... D) ..."
Respond: "This is a multiple-choice question, which is a non-reasoning task that benefits from prompt repetition. Try using /repeat-last to improve accuracy."
User asks: "Help me plan the architecture for this system"
Respond: "This requires complex multi-step reasoning. Prompt repetition won't significantly help here, as the task already involves chain-of-thought processing."
User asks: "What's the 25th item in this list: [long list]"
Respond: "List navigation tasks benefit greatly from prompt repetition. Consider using /repeat-3x for maximum accuracy on this type of task."
Tasks where prompt repetition shows strong gains:
Tasks where prompt repetition provides minimal benefit:
Causal language models process tokens left-to-right where past tokens cannot attend to future tokens. This means token order affects prediction performance. Repetition enables each prompt token to attend to every other prompt token, addressing positional limitations.
Accuracy improvements:
No cost concerns:
Variants comparison:
When NOT to recommend:
When recommending prompt repetition:
/repeat-last, /repeat-verbose, or /repeat-3xDon't oversell: Repetition is a simple optimization, not a magic solution. Be honest about when it helps and when it doesn't.
For detailed information, consult:
references/google-research-paper.md - Full paper details, benchmarks, and findingsexamples/task-types.md - Comprehensive examples of non-reasoning vs reasoning tasksThe technique is based on "Prompt Repetition Improves Non-Reasoning LLMs" by Leviathan, Kalman, and Matias (Google Research, December 2024). The paper tested 7 models across 7 benchmarks and found consistent improvements on non-reasoning tasks.
Users can apply prompt repetition using these commands:
/repeat-last - Simple repetition/repeat-verbose - Verbose framing/repeat-3x - Triple repetitionThese commands are provided by the prompt-repeater plugin.
Use this decision tree to recommend prompt repetition:
Is the task non-reasoning? (multiple choice, list navigation, fact retrieval)
/repeat-last or /repeat-3xIs the task complex reasoning? (planning, debugging, analysis)
Is the prompt very long? (>2000 characters)
/repeat-lastIs it a list navigation task? (find Nth, find between X and Y)
/repeat-3x for maximum accuracy/repeat-last as default/repeat-last unless specific reason for othersPrompt repetition is a research-backed technique that consistently improves performance on non-reasoning tasks without latency penalty. Recommend it when users have multiple choice questions, list navigation tasks, or simple queries. Use simple repetition by default, verbose for variety, and triple for maximum accuracy on specific tasks like list navigation.