INDEX

Explanations

different forms and variations of the verb "walk."

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

walk and direction/purpose

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

Negative Logits

endregion

-0.53

 particular

-0.53

 Particular

-0.45

sizePolicy

-0.43

 Aguilera

-0.43

 Dear

-0.41

 Snyder

-0.40

 Secret

-0.40

CppMethod

-0.40

 certain

-0.40

POSITIVE LOGITS

Walk

1.34

 Walk

1.26

walk

1.25

 walk

1.22

 WALK

1.20

 Walks

1.11

WALK

1.11

 walked

1.04

 walks

1.00

Walking

1.00

Activations Density 0.011%

different forms and variations of the verb "walk."

walk and direction/purpose

No Known Activations