INDEX

Explanations

instances of the word "can" and its variants, indicating possibilities or capabilities

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

can + verb

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

may

-0.37

is

-0.34

 often

-0.34

 occurs

-0.33

 awards

-0.32

 mourut

-0.31

 faptul

-0.31

 rozhod

-0.30

 might

-0.30

 actions

-0.30

POSITIVE LOGITS

 PeEnEo

0.83

TargetException

0.79

 estekak

0.76

expandindo

0.73

IVEREF

0.71

<unused74>

0.71

<unused41>

0.71

<unused47>

0.71

[@BOS@]

0.70

<unused51>

0.70

Activations Density 0.021%

instances of the word "can" and its variants, indicating possibilities or capabilities

can + verb

No Comments

No Known Activations