INDEX

Explanations

references to specific social groups or communities

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

group nouns followed by verbs

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

Их

-0.42

hermoso

-0.34

 Handlungen

-0.34

 FACILITY

-0.34

 їх

-0.34

jalá

-0.33

Билгалдахарш

-0.33

 contacts

-0.32

Pron

-0.32

FREQUENCY

-0.32

POSITIVE LOGITS

EDEFAULT

0.80

 ourselves

0.66

UnusedPrivate

0.65

rungsseite

0.60

Kjelder

0.57

atguigu

0.56

 myself

0.55

 doInBackground

0.54

mybatisplus

0.54

mobileqq

0.50

Activations Density 0.071%

references to specific social groups or communities

group nouns followed by verbs

No Comments

No Known Activations