INDEX

Explanations

statistical metrics related to diagnostic performance

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 17-gemmascope-res-65k

Configuration

google/gemma-scope-2b-pt-res/layer_17/width_65k/average_l0_125

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

65,536

Data Type

float32

Hook Name

blocks.17.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

 turbul

-0.45

гар

-0.43

 relâche

-0.42

lu

-0.41

Jump

-0.39

mt

-0.38

 Jump

-0.38

 turbulence

-0.38

kn

-0.37

sto

-0.36

POSITIVE LOGITS

 AssemblyTitle

0.87

참고

0.74

 scoring

0.73

^(@)

0.73

 scorer

0.72

 Monfieur

0.69

iseite

0.69

 myſelf

0.69

horabuena

0.68

 الحره

0.68

Activations Density 0.347%

statistical metrics related to diagnostic performance

No Comments

No Known Activations