INDEX

Explanations

phrases indicating possession or association

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

duration followed by noun

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

 bâtiment

-0.37

 trouvera

-0.36

เร็ว

-0.36

 endroits

-0.35

 caza

-0.34

 réfrig

-0.33

 touristes

-0.33

the

-0.33

 religieuse

-0.33

 Scherer

-0.33

POSITIVE LOGITS

__":

0.83

posedge

0.82

 Wikimedijinoj

0.81

__':

0.79

 мәкал

0.78

 BorderSide

0.76



0.73

 hinweg

0.72

 snippetHide

0.71

ValueStyle

0.70

Activations Density 0.019%

phrases indicating possession or association

duration followed by noun

No Comments

No Known Activations