INDEX

Explanations

references to previously established methods and studies

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B @ 20-gemmascope-res-16k

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_16k/average_l0_68

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

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

findpost

-0.48

CodeAttribute

-0.38

 GenerationType

-0.37

=$?

-0.36

ſelf

-0.36

 tričko

-0.35

UnusedPrivate

-0.34

 utafitiHapana

-0.34

 accompagné

-0.33

POSITIVE LOGITS

出版年

0.59

 previously

0.55

previously

0.54

 kasarigan

0.50

 Италијани

0.46

 norme

0.45

turo

0.44

sphase

0.44

Previously

0.43

ꉂ

0.43

Activations Density 0.902%

references to previously established methods and studies

No Comments

No Known Activations