INDEX

Explanations

the word "drop" and its variations in different contexts

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B @ 35-gemmascope-res-16k

Configuration

google/gemma-scope-9b-pt-res/layer_35/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.35.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

SequentialGroup

-0.54

RTDA

-0.50

ftagPool

-0.48

UserScript

-0.46

culada

-0.43

newBuilder

-0.43

 protoimpl

-0.42

 parl

-0.42

/*

-0.41

strSql

-0.41

POSITIVE LOGITS

 Drop

1.52

 drop

1.52

Drop

1.41

 dropped

1.40

drop

1.39

 DROP

1.36

 dropping

1.30

 drops

1.22

Dro

1.21

DROP

1.20

Activations Density 0.093%

the word "drop" and its variations in different contexts

No Known Activations