{"version":1,"type":"story","url":"https://digestai.news/story/researchers-find-instruction-exposure-unlocks-rhetorical-patterns-in-l","json":"https://digestai.news/story/researchers-find-instruction-exposure-unlocks-rhetorical-patterns-in-l.json","markdown":"https://digestai.news/story/researchers-find-instruction-exposure-unlocks-rhetorical-patterns-in-l.md","slug":"researchers-find-instruction-exposure-unlocks-rhetorical-patterns-in-l","headline":"Researchers find instruction exposure unlocks rhetorical patterns in LLMs","summary":"A new study on arXiv explores how language models absorb rhetorical strategies during pretraining, not just facts. The authors extracted an ‘Analogy Vector’ from residual streams to measure how much these models can be steered toward analogy-based reasoning. Testing seven instruction-tuned models—including Gemma-2, Llama-3.1, and Qwen-2.5—they found a 22% improvement in analogy quality for Gemma-2-9B-IT on held-out prompts, raising the harmonic mean from 3.22 to 3.94.\n\nThe research reveals a gradient of behavioral accessibility: models with instruction data woven into pretraining (like Qwen-2.5) show higher cosine similarity scores (0.94–0.97) for the analogy vector, compared to models with strict pretraining/instruction separation (Gemma-2, scoring 0.37–0.66). This challenges the binary ‘base vs. instruct’ framing, suggesting that the degree of instruction exposure during training determines how easily rhetorical patterns can be activated at inference time.","keyPoints":["Analogy Vector extracted from residual streams improves analogy reasoning by 22% for Gemma-2-9B-IT","Qwen-2.5 base models show higher cosine similarity (0.94–0.97) than Gemma-2 (0.37–0.66) due to instruction exposure","Study challenges ‘base vs. instruct’ framing, arguing instruction exposure during pretraining unlocks rhetorical patterns"],"whyItMatters":"This work could help researchers design more interpretable models by revealing how instruction exposure shapes latent rhetorical knowledge, potentially improving explainability and fine-tuning strategies.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":[],"models":["Gemma-2","Llama-3.1","Qwen-2.5"],"people":[]},"firstPublishedAt":"2026-10-30T04:00:00Z","updatedAt":"2026-10-30T04:00:00Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"arXiv cs.CL","title":"From Geometry to Behavior: How Instruction Exposure Unlocks Latent Rhetorical Directions in LLMs","url":"https://arxiv.org/abs/2610.37685","publishedAt":"2026-10-30T04:00:00Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Researchers find instruction exposure unlocks rhetorical patterns in LLMs\", 30 October 2026, https://digestai.news/story/researchers-find-instruction-exposure-unlocks-rhetorical-patterns-in-l","publisher":"Digest AI","title":"Researchers find instruction exposure unlocks rhetorical patterns in LLMs","datePublished":"2026-10-30T04:00:00Z","url":"https://digestai.news/story/researchers-find-instruction-exposure-unlocks-rhetorical-patterns-in-l"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}