INDEX

Explanations

negations and terms related to affirmation and contradiction

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

is or not

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

popd

-0.40

 trän

-0.38

 Zucchini

-0.38

LEGGI

-0.37

ucchini

-0.37

 pylint

-0.36

Hô

-0.35

 Milford

-0.34

 custodia

-0.34

POSITIVE LOGITS

是

0.89

的是

0.77

是

0.73

是小

0.70

Là

0.70

]]=

0.68

)}=

0.65

以上就是

0.64

 adalah

0.63

 является

0.63

Activations Density 0.001%

negations and terms related to affirmation and contradiction

is or not

No Comments

No Known Activations