Fireworks 用 Anthropic J-Lens 探针测试 Kimi K3 与 Qwen 3.5-9B
Fireworks 把 Anthropic 的 J-Lens 探针装到 Kimi K3 和 Qwen 3.5-9B 上，发现模型在蹦出第一个 token 前就把词准备好了。看内部状态搞可解释性，挺有意思。
Fireworks AI 将 Anthropic 的 J-Lens 探针部署到开源模型 Kimi K3 和 Qwen 3.
5-9B 上。通过检查模型的内部状态，发现它们在生成首个 token 之前就已布置好对应词汇。
相关结果发布在 Fireworks 的 J-Lens 研究博客页面。
We tested out Anthropic’s J-Lens probe at two open models: Kimi K3 and Qwen 3.5-9B. By looking at their internal states, we found that the models are setting up the right vocabulary before a single token hits the page. Check out the research: fireworks.ai/blog/J-Lens-Ki…📊 2 ⚡
This report is categorized under AITOP's "Papers" desk, focusing on China's AI industry and domestic foundation model ecosystem. Key entities and technical topics under continuous tracking include J-Lens, Kimi K3, Qwen 3.5-9B, Anthropic.
Source attribution: curated from Fireworks AI (@FireworksAI_HQ). Cross-source corroboration: 11 independent sources have covered this development (including lmarena.ai, The Rundown AI, elvis), confirming industry-wide signal strength.