INDEX

Explanations

patterns in numerical data or sequences

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

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Harg

-0.40

shadowOpacity

-0.39

Fla

-0.39

qrst

-0.39

MAR

-0.39

ple

-0.37

<0xED>

-0.37

 Brow

-0.37

fum

-0.36

POSITIVE LOGITS

 enfans

0.76

 ainfi

0.72

ๆ

0.70

 Sünde

0.65

bibfield

0.65

 dezelve

0.64

 Geiſt

0.63

 feroit

0.63

 Juifs

0.62

 Comprometido

0.61

Activations Density 0.163%

patterns in numerical data or sequences

No Comments

No Known Activations