Brief

Researchers have developed a new benchmark called Arbitrary Preference Mapping (APM) to evaluate how well large language models can adapt to users' implicit style preferences. The APM benchmark uses a randomized mapping to decouple user attributes from response principles, preventing models from relying on stereotypes and forcing them to infer preferences from conversation history. Experiments using this methodology on Llama-3.1-8B and Qwen-3.5-27B showed that routing-based personalization methods were the most effective, while other approaches like RAG and soft prompt optimization showed limited improvement. AI

Researchers have developed TorchSight, an open-source local system for classifying security documents using a fine-tuned Qwen 3.5 27B large language model. This system achieved 95.0% accuracy on a benchmark of 1,000 documents, significantly outperforming commercial models which scored between 75.4% and 79.9%. The fine-tuned local model demonstrates the capability to maintain data privacy while accurately identifying sensitive information across various security categories and subcategories. AI