INDEX

Explanations

terms related to "Mal" or "malicious" concepts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 21-gemmascope-res-16k

Configuration

google/gemma-scope-2b-pt-res/layer_21/width_16k/average_l0_70

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.21.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

 createState

-0.77

 signé

-0.58

alyptus

-0.57

 Clermont

-0.57

curities

-0.55

 trouvait

-0.55

RetentionPolicy

-0.55

 écoles

-0.54

 préfé

-0.52

++]=

-0.52

POSITIVE LOGITS

Mal

0.87

Mal

0.84

MAL

0.82

mal

0.71

MAL

0.69

tvguidetime

0.63

mal

0.63

richTextPanel

0.61

 Malt

0.58

+:+

0.57

Activations Density 0.078%

terms related to "Mal" or "malicious" concepts

No Comments

No Known Activations