INDEX

Explanations

keywords related to legal and procedural obligations

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

survive intermediate scrutiny

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

InjectMocks

-0.73

 NSCoder

-0.68

 autorytatywna

-0.67



-0.67

.*")]

-0.66

المناصب

-0.66

niſſe

-0.65

tvguidetime

-0.64

 Weiſe

-0.64

 dieſes

-0.63

POSITIVE LOGITS

通

0.37

 deten

0.30

런

0.29

通

0.28

κ

0.28

fscanf

0.27

Msgs

0.27

 hins

0.25

By

0.25

hithe

0.25

Activations Density 0.233%

keywords related to legal and procedural obligations

survive intermediate scrutiny

No Comments

No Known Activations