INDEX

Explanations

phrases related to actions of doing or having done something

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

end of phrase

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

 Jerusalén

-0.49

UpInside

-0.48

épreuve

-0.48

Totale

-0.47

 Aspekte

-0.45

最快更新

-0.45

kuuta

-0.45

 muchísimo

-0.44

 libros

-0.43

Hozzáférés

-0.43

POSITIVE LOGITS

 ujednoznacz

0.43

 FetchType

0.40

Chief

0.39

 betweenstory

0.39

 Chief

0.39

volley

0.38

 HttpNotFound

0.38

 running

0.38

下载附件

0.37

StructEnd

0.36

Activations Density 0.052%

phrases related to actions of doing or having done something

end of phrase

No Comments

No Known Activations