做工具调用智能体开发的团队，如果遇到模型明明知道该用工具却就是不调用的怪现象，这篇论文直接点出了后层几何结构的根本原因，值得一读。
一篇关于工具使用智能体的可解释性论文揭示了模型在认知与行动之间的脱节：模型内部状态显示它知道应该调用工具，但在实际输出中却未能执行。这种不匹配率在26%到54%之间，且完全集中在认知到行动的转换阶段，而非认知本身。研究发现，模型内部的方向是可解码的，但后层最后一个token的表示几乎与产生的行动正交，导致信号丢失。该工作试图预测哪些干预措施有效，哪些无效。对于在工具调用提示上做A/B测试却遇到奇怪上限的开发者，这篇论文可能提供了很好的解释。
Interesting interpretability paper on tool-using agents. The authors probe hidden states and find t...
Interesting interpretability paper on tool-using agents. The authors probe hidden states and find the model often recognizes it should call a tool, but fails to actually call one. The mismatch ranges from 26 to 54%, and it concentrates entirely in the cognition-to-action transition, not in cognition itself. In other words, the model usually knows it should call the tool. The internal probe direction is decodable. But the late-layer last-token regime rotates that signal nearly orthogonal to the action it produces. This work tries to predict which interventions will actually work and which will not. Most will blame bad prompting or weak tool-call training, and probably ignore the late-layer geometry. If you have been A/B testing tool-use prompts and getting weird ceilings, this work might offer a good explanation to that behavior. Paper: arxiv.org/abs/2605.14038 Learn to build effective AI agents in our academy: academy.dair.ai 💬 5 🔄 5 ❤️ 17 👀 2206 📊 10 ⚡