INDEX

Explanations

terms related to ninjas and assassins

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

 Verhältnisse

-0.68

⟬

-0.65

))/(

-0.61

SequentialGroup

-0.60

]")]

-0.59

 Verhält

-0.59

 bedoeld

-0.59

 Hebron

-0.58

rungsseite

-0.58

]}>

-0.57

POSITIVE LOGITS

 ninja

0.99

 Ninja

0.98

 ninjas

0.93

Ninja

0.90

ninja

0.82

Nin

0.75

忍者

0.65

nin

0.58

🥷

0.56

NIN

0.55

Activations Density 0.019%

terms related to ninjas and assassins

No Comments

No Known Activations