LLM Linear-Probe Activation-Steering Pipeline Diagram

LLM Linear-Probe Activation-Steering Pipeline Diagram

Overview

A three-stage pipeline diagram of steering a language model with a linear probe. It is an editable PowerPoint diagram adapted from the original paper figure under CC BY 4.0.

What the diagram shows

Extraction: N sample documents pass through Layer 1 … Layer i … Layer L, and the last-token embedding hi of a hidden-state row is collected into a colour-coded hidden matrix H (N × P). Training: a Standardize card X = (H − μ) / s feeds a logistic-regression solver card with a small accuracy-vs-λ curve and magnifier, and the objective minw ℒ + λ‖w‖22. A Rescale & Norm card gives ŵ = w / s and v = ŵ / ‖ŵ‖2, producing the concept vector v (P × 1). Inference steering: the update h ← h + αv is added to the last-token state hi of a user prompt (chat-bubble icon) inside the same layer stack. Every box, arrow, label and math term is a native shape you can recolor, resize and relabel for your own slides, posters or papers.

File format

The download contains .pptx (1 slide, 15.58 × 6.28 in; opens in PowerPoint, Keynote, WPS and LibreOffice Impress; Google Slides works too, but complex freeforms may shift slightly), .svg and .drawio. Includes preview.png (full resolution) and a 1200 px web preview.

Shapes and editing

The file has 139 native editable shapes: 56 auto-shapes, 36 text boxes, 47 freeforms, 0 lines/connectors. No embedded images. Everything is vector. You can edit all text and math (hi, H, X = (H − μ) / s, λ, minw ℒ + λ‖w‖22, ŵ = w / s, v = ŵ / ‖ŵ‖2, h ← h + αv), every layer box and dotted ellipsis, the hidden-state token rows, every cell of the hidden matrix, the Standardize, solver and Rescale & Norm cards, the accuracy curve with axes and magnifier icon, the concept-vector bar, the document and chat-bubble icons and all elbow connectors.

Embedded images and fonts

This file has no embedded images; everything is vector. Body text uses Calibri and formulas use Times New Roman. Calibri ships with Microsoft Office and Times New Roman is preinstalled on Windows and macOS (LibreOffice substitutes the metric-compatible Carlito and Liberation Serif). The script ℒ uses Cambria Math, which ships with Microsoft Office; other apps substitute a math font. Sub- and superscripts are real formatted text runs, not equation objects. The SVG keeps text as text; the draw.io file is a shape-by-shape conversion and may need small nudges. This is a faithful redraw for reuse, not the authors' original source file.

Source & license Adapted from Figure 1 of RAPTOR: Ridge-Adaptive Logistic Probes by Ziqi Gao, Yaotian Zhu, Qingcheng Zeng, Xu Zhao, Ziqing Wang, Feng Ruan, Kaize Ding (arXiv:2602.00158, https://arxiv.org/abs/2602.00158). Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Redrawn as editable shapes; changes made by 叫我Alonzo就好了/alonzolab. Not endorsed by the original authors.

Questions or feedback? Email [email protected]