“Often the math works out that you get a lot more for $20 a month if you settle for smaller sized but capable models (8b-30b). I don’t see how it’s better other than Ollama can “promise” they don’t store your data where as OpenRouter is dependent on which host you choose (and there’s no indicator on OpenRouter exposing which ones do or don’t). In a univer…”
Encryption-Friendly LLM Architecture
Large language models (LLMs) offer personalized responses based on user interactions, but this use case raises serious privacy concerns. Homomorphic encryption (HE) is a cryptographic protocol supporting arithmetic computations in encrypted states and provides a potential solution for privacy-preserving machine learning (PPML). However, the computational intensity of transformers poses challenges for applying HE to LLMs. In this work, we propose a modified HE-friendly transformer architecture with an emphasis on inference following personalized (private) fine-tuning. Utilizing LoRA fine-tuning and Gaussian kernels, we achieve significant computational speedups -- 6.94x for fine-tuning and 2.3x for inference -- while maintaining performance comparable to plaintext models. Our findings provide a viable proof of concept for offering privacy-preserving LLM services in areas where data protection is crucial. Our code is available on GitHub.
候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。
HN 书评
推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及“LLMs are at the current forefront of FHE research. There are a few papers doing some tweaked versions of BERT in <1 minute per token. Which is only ~4 orders of magnitude slower than cleartext. https://arxiv.org/html/2410.02486v1#S5”