INDEX

Explanations

proper nouns or names in various contexts

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

numbers and specific words

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

 ervoor

-0.40

IRQn

-0.34

 scio

-0.34

ießen

-0.34

 vực

-0.31

Outlined

-0.30

ÍN

-0.30

and

-0.30

clear

-0.29

POSITIVE LOGITS

 ब्रेकडाउन

0.79

OGND

0.73

хьтан

0.71

 Infórmanos

0.69

GEBURTSDATUM

0.69

UserScript

0.69

Vidite

0.68

RenderAtEndOf

0.67

StructEnd

0.67

ValueStyle

0.66

Activations Density 0.234%

proper nouns or names in various contexts

numbers and specific words

No Comments

No Known Activations