INDEX

Explanations

instances of the word "before" or phrases indicating prior events

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B @ 36-gemmascope-res-16k

Configuration

google/gemma-scope-9b-pt-res/layer_36/width_16k/average_l0_120

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.36.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

jenigen

-0.46

 olmayan

-0.46

 tidigare

-0.36

 pamoja

-0.35

envolvimento

-0.35

 näky

-0.35

くない

-0.34

earlier

-0.34

‍♀️

-0.34

 ayrıca

-0.34

POSITIVE LOGITS

hand

1.00

HAND

0.72

 anything

0.66

andafter

0.63

 anyone

0.62

 hand

0.61

 they

0.60

transQ

0.60

CascadeType

0.59

anything

0.59

Activations Density 0.108%

instances of the word "before" or phrases indicating prior events

No Comments

No Known Activations