Elvis Saravia：Anthropic 模型的拒绝机制阻碍自主科研循环
做自主科研循环的朋友可以看看这条，Saravia 用亲身经验说 Anthropic 模型的拒绝机制会频繁打断 autoresearch 流程，连论文复现都经常失败。如果你的工作流涉及改进 AI 系统本身，选模型前要掂量一下 Claude 的拒答问题，可能 GPT 或 Gemini 这类限制更松的模型更合适。
AI研究者 Elvis Saravia（omarsar0）在 X 上发帖称，Anthropic 有权自由处置自己的模型，但科研人员在用这些模型推进开放科学时会因拒绝机制遭遇很大困难。他基于个人经验指出，问题不只是论文复现（多数情况下也会失败），而是用于改进 AI 系统某些方面的 autoresearch 循环（自主研究循环）几乎无法在这些模型上运行。该帖发布后获得约 800 次浏览和 3 次点赞。
Saravia 的观点针对的是 Anthropic 模型以安全对齐严格著称的拒绝行为：当模型被用于自动化研究工作流时，涉及 AI 系统改进本身的提示会频繁触发拒答，导致自主循环中断。他强调自己并非指责 Anthropic，而是提醒以科研为目的的用户在选择模型时需要考虑这一实际限制。
Nothing against Anthropic. They are free to decide whatever they want with their models. I mention RSI because if you are a scientist trying to advance open science with these models, you will have a really rough time with the refusals. I can talk from personal experience. And I am not talking about paper reproductions (even that fails in most cases). I mean autoresearch loops to improve certain aspects of AI systems.📊 1 ⚡
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Anthropic, Claude, 科研工作流, AI安全.
Source attribution: curated from elvis (@omarsar0).
