INDEX

Explanations

phrases that indicate generating or proposing ideas

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

come up with ideas

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

 Doors

-0.58

transQ

-0.54

Sunlight

-0.52

Doors

-0.51

RDONLY

-0.50

DataBinding

-0.49

 Safely

-0.49

(*(

-0.49

gridx

-0.48

 DataAccess

-0.48

POSITIVE LOGITS

 придума

0.74

 devised

0.74

 ideas

0.71

 idea

0.65

 devise

0.63

 conco

0.59

 inventive

0.58

 devising

0.57

 invented

0.57

 elaborate

0.56

Activations Density 0.009%

phrases that indicate generating or proposing ideas

come up with ideas

No Comments

No Known Activations