← 研踪 PaperTrailarXiv:2611.13322 · v1
MicroLens: Read, Trace, and Rewire the Residual Stream with Two Lenses
首次提交 2026-11-10 06:43
版本更新 2026-11-10 06:43
雷达观测 2026-11-11 11:04
摘要
Lenses that decode hidden states into words are a popular way to look inside language models, but a heatmap of decoded words only shows what a vector reads as and cannot distinguish whether, or how much, the model uses the information. We present an interactive viewer that puts reading, tracing and editing on the same layers-by-tokens grid of the residual stream. In the viewer, every cell can be decoded with two lenses, the logit lens and the Jacobian lens (J-lens), traced back to the tokens it drew from, and edited, with the rest of the model rerun on the spot. We demonstrate three usage scenarios on Gemma-3-4B-it, backed by sweeps over a few hundred prompts, to show what this tool can discover. First, the two lenses give different directions to write with, each suiting a different edit: the logit lens's word direction sets what a token outputs, while the J-lens's direction changes what later tokens read. Second, in two-hop trivia questions (the capital of the state or country containing a given city), the state or country is readable at the city token from as early as layer 7, and swapping it there can redirect the answer 92-98% of the time; however, at the answer position, the state or country information is not used. Third, when a chat model answers questions with a step-by-step reasoning prompt, the correct answer is often already the top word in the residual stream before any reasoning is written: for 89-92% of retrieval questions and 39-63% of multi-hop and serial ones, far above their respective distractors (at most 5%). However, ordering puzzles are an exception, as the model works them out only once it writes its final reasoning steps. Furthermore, for answers that are visible in the residual stream, the J-lens reads them a layer or two earlier than the logit lens across the different types of question.
作者备注:9 pages, 11 figures. Code: https://github.com/Ni2002ka/residual_stream_vis/tree/v1.0-pacificvis
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-10 06:43 · 最新观测版本
MicroLens: Read, Trace, and Rewire the Residual Stream with Two Lenses
Lenses that decode hidden states into words are a popular way to look inside language models, but a heatmap of decoded words only shows what a vector reads as and cannot distinguish whether, or how much, the model uses the information. We present an interactive viewer that puts reading, tracing and editing on the same layers-by-tokens grid of the residual stream. In the viewer, every cell can be decoded with two lenses, the logit lens and the Jacobian lens (J-lens), traced back to the tokens it drew from, and edited, with the rest of the model rerun on the spot. We demonstrate three usage scenarios on Gemma-3-4B-it, backed by sweeps over a few hundred prompts, to show what this tool can discover. First, the two lenses give different directions to write with, each suiting a different edit: the logit lens's word direction sets what a token outputs, while the J-lens's direction changes what later tokens read. Second, in two-hop trivia questions (the capital of the state or country containing a given city), the state or country is readable at the city token from as early as layer 7, and swapping it there can redirect the answer 92-98% of the time; however, at the answer position, the state or country information is not used. Third, when a chat model answers questions with a step-by-step reasoning prompt, the correct answer is often already the top word in the residual stream before any reasoning is written: for 89-92% of retrieval questions and 39-63% of multi-hop and serial ones, far above their respective distractors (at most 5%). However, ordering puzzles are an exception, as the model works them out only once it writes its final reasoning steps. Furthermore, for answers that are visible in the residual stream, the J-lens reads them a layer or two earlier than the logit lens across the different types of question.
查看此版本解读 →查看此版本原文 ↗