INDEX

Explanations

verbs and phrases indicating preference or recommendation

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B @ 1-gemmascope-res-16k

Configuration

google/gemma-scope-9b-pt-res/layer_1/width_16k/average_l0_69

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.1.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

 μην

-0.52

 itſelf

-0.49

 Pflanze

-0.49

MemoryWarning

-0.48

 dianteiro

-0.47

 varandra

-0.45

 zostanie

-0.44

 pazi

-0.43

 muere

-0.43

 vedno

-0.42

POSITIVE LOGITS

 would

0.71

would

0.62

 Would

0.60

Would

0.58

↵

0.54

 WOULD

0.52

 wanted

0.50

Expected

0.48

 Expected

0.48

Wanted

0.48

Activations Density 0.217%

verbs and phrases indicating preference or recommendation

No Comments

No Known Activations