INDEX

Explanations

terms related to safety and security

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

safe and secure

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_24/width_131k/average_l0_97

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.24.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

 wireType

-0.51

ufficio

-0.50

 matéria

-0.45

prache

-0.44

ellem

-0.44

Ufficio

-0.42

 mídia

-0.42

 gärna

-0.41

 StringTokenizer

-0.41

abstractmethod

-0.41

POSITIVE LOGITS

 safety

1.00

 safe

1.00

 Safe

0.97

Safe

0.96

Safety

0.95

safety

0.94

safe

0.94

 Safety

0.90

 safer

0.88

 SAFE

0.87

Activations Density 0.022%

terms related to safety and security

safe and secure

No Comments

No Known Activations