用小代理模型精准预测LLM推理性能，省钱省算力的高效新方案。
arXiv:2509.21013v4 Announce Type: replace-cross Abstract: Given the prohibitive cost of pre-training large language models, it is essential to leverag…
微软开源BenchPress，无需跑分即可预测大模型在任意基准上的表现
Article URL: https://microsoft.github.io/benchpress/ Comments URL: https://news.ycombinator.com/item?id=48654705 Points: 3 # Comments: 0
一款能预测语音AI性能的音频洞察工具，帮你优化声音交互体验。
Audio insight that predicts voice AI performance Discussion | Link
探索LLM重排序器如何自我预测排序性能，挑战模型自省边界的新研究。
arXiv:2606.03535v1 Announce Type: cross Abstract: Retrieval effectiveness varies substantially across queries, making it important to estimate ranking…
提出基于示例预测LLM符号与提示程序性能的新方法，为智能编程提供高效评估框架。
arXiv:2605.21515v1 Announce Type: cross Abstract: LLM prompting is widely used for naturally stated tasks, yet it is unreliable it may succeed on a fe…
融合布局几何与网表拓扑，跨注意力机制精准预测标准单元性能，加速芯片设计流程。
arXiv:2605.20287v1 Announce Type: new Abstract: Standard cells form the building blocks of digital circuits, so their delay and power critically influ…
用代理指标提前预判LLM下游表现，为模型选型提供可靠决策依据
arXiv:2605.18607v1 Announce Type: cross Abstract: Progress in language model development is often driven by comparative decisions: which architecture …