INDEX

Explanations

phrases that indicate reasons or justifications related to actions

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

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 plazos

-0.43

 agrí

-0.42

 bientôt

-0.41

MethodManager

-0.39

 discusión

-0.39

 dispozici

-0.39

 menac

-0.38

싶

-0.38

wiſe

-0.37

 academy

-0.36

POSITIVE LOGITS

 colorWith

0.57

 having

0.55

 stringBy

0.55

Parcelize

0.55

Наводи

0.54

vì

0.52

BorderFactory

0.49

yarnpkg

0.48

 doing

0.48

 being

0.47

Activations Density 0.037%

phrases that indicate reasons or justifications related to actions

No Comments

No Known Activations