INDEX

Explanations

references to wood and related terms

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

 ubezpiec

-0.42

TagMode

-0.41

ithub

-0.39

 málo

-0.39

 arşivlendi

-0.39

 allAfrica

-0.39

tabak

-0.38

 ripet

-0.38

 lauk

-0.37

RenderAtEndOf

-0.37

POSITIVE LOGITS

 Wood

0.57

pecker

0.56

Wood

0.56

pile

0.56

 chips

0.54

grain

0.52

chips

0.51

 Chips

0.50

henge

0.48

ModelAndView

0.47

Activations Density 0.075%

references to wood and related terms

No Comments

No Known Activations