Anthropic推LLM可证明推理验证
推荐指数 80.0 NO. 017 · 2026.11.05
发布2026/11/04Score157Comments77
简介
Anthropic联合MIT、EPFL等机构提出了一种密码学方案，让用户能密码学验证AI回复确实来自声称的模型，而非被中间商偷换的廉价替代或篡改输出。这对AI API托管和模型即服务赛道的信任机制有根本性影响。
编辑判断
当前模型API市场最大的灰色地带是'路由欺诈'——中间商把用户请求偷偷转给更便宜的模型，用户靠输出质量根本分辨不出来。Anthropic这篇工作用SNARKs把模型推理过程变成可验证的密码学证明，理论上能彻底堵死这个漏洞。
但工程落地还有两道坎：一是证明生成开销目前比推理本身高几个数量级，二是需要模型架构层面的配合，闭源模型如GPT-4不会接这个标准。最可能先落地的场景是金融、医疗等强合规领域的私有化部署，以及去中心化AI推理网络（如Ritual、Gensyn这类项目）作为差异化卖点。
做AI基础设施或安全合规方向的团队值得跟踪，如果证明开销能降到10倍以内，这会成为模型托管的标配。
社区反馈
意见分歧 71 条评论
核心争论：密码学验证模型身份是刚需还是伪需求，能否解决用户真正关心的输出质量与信任问题
> A user querying a model today cannot check that the responses came from the specific model that a developer claims is responding Indeed, I wonder what kind of developer would make users worry about that...? Funny PR stunt coming from a company that was in the news for silently routing ML-related r
I've read that providers like those you find on OpenRouter can sometimes route to other models; this would catch that. But I've also heard, as complete rumour, sometimes some named models perform significantly worse on some hosts and the theory is they're running quantized, without that being publis
Maybe it's possible to tell by looking at the size of that error? > The integer version is generally expected and empirically tested to be within some small error rate of true floating-point inference generically (even taking into account non-determinism of floating-point computations), though this