INDEX

Explanations

references to user-related data or file paths

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

file paths and directory structures

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

 yapılan

-0.43

 zví

-0.40

 blitt

-0.38

 refroid

-0.37

 almacenar

-0.37

 conseils

-0.37

 recettes

-0.36

 patrimoine

-0.36

 yaptığı

-0.35

 régler

-0.35

POSITIVE LOGITS

fjspx

0.64

Diweddarwch

0.64

 Infórmanos

0.63

 Taktlose

0.63

Autoritní

0.63

SequentialGroup

0.62

intios

0.59

 للاسماء

0.57

ंदीखरीदारी

0.57

GraphicsUnit

0.56

Activations Density 0.002%

references to user-related data or file paths

file paths and directory structures

No Comments

No Known Activations