INDEX

Explanations

references to individuals and their past roles or statuses

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

words describing a previous professional status or role (particularly "former" followed by a professional title).

oai_token-act-pair · claude-3-5-haiku-20241022 Triggered by @neilrathi

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

astify

-0.60

뀜

-0.48

뀝

-0.45

DockStyle

-0.45

 contemporaine

-0.44

hæng

-0.44

 samfun

-0.43

contentLoaded

-0.43

 næste

-0.43

guenos

-0.43

POSITIVE LOGITS

 former

0.87

 formerly

0.81

 Former

0.78

Former

0.71

 Formerly

0.69

former

0.64

 früher

0.64

formerly

0.63

 previous

0.63

 früheren

0.63

Activations Density 0.289%

references to individuals and their past roles or statuses

words describing a previous professional status or role (particularly "former" followed by a professional title).

No Comments

No Known Activations