INDEX

Explanations

variations of the word "mode."

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

mode of

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

The

-0.51

 Públicas

-0.50

 colgantes

-0.49

 pelúcia

-0.48

Gesam

-0.48

 calcetines

-0.46

Fernseh

-0.45

 amarillas

-0.44

 científicas

-0.44

 vérification

-0.43

POSITIVE LOGITS

 Mode

1.31

mode

1.30

 MODE

1.28

 mode

1.28

MODE

1.27

Mode

1.24

setMode

1.11

 modes

1.09

 Modes

1.02

modes

1.02

Activations Density 0.018%

variations of the word "mode."

mode of

No Comments

No Known Activations