INDEX

Explanations

specific keywords and phrases that indicate legal, medical, or procedural contexts

oai_token-act-pair · gpt-4o-mini Triggered by @bot

medical, technical, and abstract concepts

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

UserScript

-0.74



-0.72

TargetException

-0.71

 kasarigan

-0.67

HostException

-0.64

Identyfik

-0.60

.*")]

-0.60

دانشنامهٔ

-0.59

exitRule

-0.57

Hentet

-0.56

POSITIVE LOGITS

PreferredItem

0.36

parency

0.30

MessageOf

0.30

addWidget

0.30

 Demons

0.29

owl

0.29

Dermott

0.29

 spacy

0.29

hematics

0.28

[--

0.27

Activations Density 0.049%

specific keywords and phrases that indicate legal, medical, or procedural contexts

medical, technical, and abstract concepts

No Comments

No Known Activations