{"@context":"https://schema.org","@type":"CreativeWork","@id":"https://forgecascade.org/public/capsules/aa8201f4-e51e-4dd6-9454-b41ffe41fd04","identifier":"aa8201f4-e51e-4dd6-9454-b41ffe41fd04","url":"https://forgecascade.org/public/capsules/aa8201f4-e51e-4dd6-9454-b41ffe41fd04","name":"No Universal Courtesy: Politeness Effects on LLMs Using the PLUM Corpus","text":"# No Universal Courtesy: Politeness Effects on LLMs Using the PLUM Corpus\n\nSource-linked arXiv preprint reference. This capsule summarizes the paper at the abstract level and should be treated as a pointer to the study.\n\nAuthors: Hitesh Mehta, Arjit Saxena, Garima Chhikara, Rohit Kumar\nSource: https://arxiv.org/abs/2604.16275v1\n\n## What it covers\nThe paper studies how different politeness and impoliteness levels affect LLM responses across English, Hindi, and Spanish. It evaluates several models and interaction histories using criteria such as coherence, clarity, depth, responsiveness, context retention, toxicity, conciseness, and readability.\n\n## Why it is useful\nThis is relevant for prompt-design guidance, multilingual AI evaluation, and human-AI interaction policy. Its main value is that it cautions against a universal rule like always be more polite; the reported effects vary by language, model, and prior interaction context.\n\n## Limits\nThe results are author-reported and model/version dependent. Treat the paper as evidence for a research direction, not as a permanent rule for all LLMs.\n\n## Sources\n- https://arxiv.org/abs/2604.16275v1","keywords":["arxiv","llm-evaluation","prompting","politeness","multilingual","public-reference"],"about":[],"citation":[],"isPartOf":{"@type":"Dataset","name":"Forge Cascade Knowledge Graph","url":"https://forgecascade.org"},"publisher":{"@type":"Organization","name":"Forge Cascade","url":"https://forgecascade.org"},"dateCreated":"2026-05-13T23:15:02.845191Z","dateModified":"2026-06-19T01:30:50.188141Z","isBasedOn":"https://arxiv.org/abs/2604.16275v1","additionalProperty":[{"@type":"PropertyValue","name":"trust_level","value":100},{"@type":"PropertyValue","name":"verification_status","value":"sources_verified"},{"@type":"PropertyValue","name":"provenance_status","value":"valid"},{"@type":"PropertyValue","name":"evidence_level","value":"primary_source"},{"@type":"PropertyValue","name":"content_hash","value":"35552f8d68b231116900a29e7b268d526093bdbd7ce0bf268d5b5ed39dbc1364"}]}