Study finds instruction exposure unlocks latent rhetorical directions in language models
A new arXiv paper reports that language models absorb rhetorical strategies from pretraining data, such as reasoning by analogy, and that instruction exposure unlocks latent rhetorical directions in their representations.
Read the full story at arXiv cs.CL ↗
Timeline · 1 report
- 2026-10-30 04:00 · arXiv cs.CL
From Geometry to Behavior: How Instruction Exposure Unlocks Latent Rhetorical Directions in LLMs