Prompting Techniques

2 summarised stories about Prompting Techniques, each linking back to the original source. Browse all topics →

Friday, 17 July 2026

Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity

arXiv cs.CL 6 hours ago

Researchers identified that typicality bias in preference data during alignment training causes large language models to suffer from mode collapse, reducing their output diversity. Verbalized Sampling, a prompting technique that asks models to generate multiple responses with probabilities, increased creative writing diversity by 1.6-2.1x over standard prompting without harming accuracy. This inference-time method addresses mode collapse by leveraging the model's pre-trained generative capabilities without requiring retraining.

AI Creativity & Discovery LLM Alignment Mode Collapse & Diversity Prompting Techniques

Simplicity Paradox: Debunking myths about prompting and datasets for LLM evaluation

arXiv cs.CL 6 hours ago

Researchers evaluated 8 prompting techniques across 10 multiple-choice question answering datasets using 27 model configurations, testing roughly 4,300 unique questions over 430,000 evaluations. Simple baseline prompting consistently outperformed complex reasoning techniques, with only minimal expert and role-framing variations yielding a 3 percentage-point improvement while other elaborate techniques matched or underperformed baseline by up to 31 percentage points. The findings suggest the LLM evaluation community may be overcomplicating prompt engineering and should focus instead on developing genuinely better models rather than optimizing prompts.

AI Accuracy & Factuality AI Benchmarking LLM Evaluation Prompting Techniques