INDEX

Explanations

terms related to legal consequences and imprisonment

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

prison sentences

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

expandindo

-0.71

Personendaten

-0.53

nambuco

-0.52

Panamoan

-0.51

معلومات

-0.47

oredCriteria

-0.47

锈钢

-0.47

iettes

-0.47

 postulate

-0.44

 confirmations

-0.44

POSITIVE LOGITS

 jail

0.40

 imprisonment

0.39

 depending

0.37

*}$

0.36

IOR

0.36

InputBorder

0.35

Jail

0.35

OrNil

0.33

 Depending

0.33

OR

0.33

Activations Density 0.014%

terms related to legal consequences and imprisonment

prison sentences

No Comments

No Known Activations