LLM translations introduce model-specific statistically significant emotional fingerprints that limit preservation of author voice, with post-editing providing partial alignment to human norms.
On the creativity of large language models
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
years
2026 4representative citing papers
GPT-4o produces humorous and relevant cartoon captions in many cases but frequently misses irony, cultural references, or narrative constraints, supporting its use as an ideation assistant in hybrid human-AI workflows.
citing papers explorer
No citing papers match the current filters.