Brief

A study of Anthropic's prompt caching on real production traffic revealed significant cost savings, with the provider's built-in caching being the most effective layer. The analysis, conducted over 330 LLM calls for AI search visibility monitoring, found that exact-match caching yielded under 5% hit rates and minimal savings, primarily serving as an idempotency feature. Semantic caching showed a higher hit rate but incurred substantial infrastructure costs, making it viable only for large-scale operations. AI

VITUR公司发布了其新款XR眼镜VITURE Beast，旨在解决XR行业长期存在的“够大、够亮、够灵”的用户痛点。该产品采用了新的Prism光波导光学方案，将视场角提升至58度，并配备了1250尼特的Micro-OLED屏幕和实时渲染补偿算法，以及原生3DoF能力，以提供更沉浸和稳定的视觉体验。VITUR公司CEO姜公略表示，这款产品不仅面向游戏和观影用户，更将办公场景作为突破口，希望通过提供移动办公和多任务处理的解决方案，推动XR眼镜走出“极客玩具”的定位，成为日常高频使用的生产力工具。 AI

Researchers have developed PRISM, a novel method for efficient fine-tuning of large language models by prioritizing data samples that most effectively guide the model toward a desired behavior. Unlike previous approaches that treat all target examples equally, PRISM weights these examples based on the current model's preference, creating a more precise target representation. This allows PRISM to concentrate the training budget on the most impactful data, leading to improved performance in both general fine-tuning and safety-oriented tasks. AI