Brief

A new research paper compares the performance of large language models (LLMs) against fine-tuned RoBERTa models for extracting complex circumstances from death investigation narratives. The study introduces a "Complexity Score" algorithm to determine optimal prompting strategies, finding that LLMs excel at low-prevalence circumstances where fine-tuned models lack sufficient training data. The research demonstrates consistent performance patterns across frontier LLMs like GPT-5.2, Gemini 2.5 Pro, and Llama-3 70B, suggesting a hybrid architecture where LLMs handle rare cases and fine-tuned models manage common ones. AI

A new study evaluated AI reviewers against human experts in assessing scientific papers, finding that AI models like GPT-5.2, Gemini 3.0 Pro, and Claude Opus 4.5 can outperform top human reviewers on certain metrics. While AI reviewers identified unique issues and were rated highly for correctness and evidence, they also exhibited weaknesses such as limited subfield knowledge and excessive overlap in their critiques. The research concludes that current AI reviewers are best utilized as complements to human expertise rather than replacements. AI

Researchers have developed OGCaReBench, a new benchmark designed to evaluate how well large language models can answer complex clinical questions that fall outside standard medical guidelines. The benchmark, derived from medical case reports and validated by experts, focuses on free-form, retrieval-based reasoning for rare scenarios. Experiments showed that even advanced models like GPT-5.2 struggled, but augmenting them with retrieved medical articles significantly improved performance, highlighting the need for evidence-grounding in medical AI. AI