INDEX

Explanations

expressions of personal favorites or highlights related to experiences or opinions

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

favorite parts or standout dishes

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

المناصب

-0.48

RTSN

-0.48

xtext

-0.47

 levis

-0.45

 ProductService

-0.45

хьтан

-0.45

RectangleBorder

-0.44

 Италијани

-0.44

parsedMessage

-0.43

THX

-0.42

POSITIVE LOGITS

 standout

0.52

 highlights

0.47

 highlight

0.46

richTextPanel

0.46

 destaque

0.43

 destacar

0.42

 favorite

0.42

 favorita

0.42

 محفوظة

0.41

 particularly

0.41

Activations Density 0.015%

expressions of personal favorites or highlights related to experiences or opinions

favorite parts or standout dishes

No Comments

No Known Activations