跨模型代理剪枝巧妙兼顾低延迟与高精度，解决长上下文LLM推理中KV缓存内存墙难题
arXiv:2605.16360v1 Announce Type: new Abstract: Efficient long-context inference in Large Language Models (LLMs) is severely constrained by the Key-Va…
提出基于校准不确定度的LLM级联路由方案，在保持性能的同时降低推理成本。
arXiv:2605.18796v1 Announce Type: new Abstract: LLM cascades and model routing promise lower inference cost by sending easy queries to a small model a…
论文揭示多模态推理中潜伏视觉令牌的非必需性，随机噪声替代不影响性能，挑战现有认知。
arXiv:2605.18641v1 Announce Type: new Abstract: Latent visual reasoning involves visual evidence more directly in multimodal reasoning by inserting co…
论文提出早停链式思维方法，减少大模型推理成本，无需白盒干预。
arXiv:2509.14004v2 Announce Type: replace Abstract: Reasoning large language models (LLMs) have demonstrated superior capacities in solving complicate…
新研究提出评测框架，分解LLM推理中token效率与准确性的关系，揭示模型是否真正高效推理而非空转。
arXiv:2602.09805v2 Announce Type: replace-cross Abstract: As reasoning LLMs increasingly trade tokens for accuracy through deliberation, search, and s…
极端压缩链式思考，大幅提升大模型推理效率，减少计算开销，创新方法值得关注。
arXiv:2602.08324v3 Announce Type: replace Abstract: Chain-of-Thought (CoT) reasoning successfully enhances the reasoning capabilities of Large Languag…
实测对比：本地M5 MacBook Pro运行LLM每百万token成本约$1.5，而OpenRouter同类模型价格仅1/3且速度翻倍，揭示本地推理的真实经济账。
Article URL: https://www.williamangel.net/blog/2026/05/17/offline-llm-energy-use.html Comments URL: https://news.ycombinator.com/item?id=48168198 Poin…