INDEX

Explanations

terms related to physical destruction or violence

oai_token-act-pair · gpt-3.5-turbo

instances of the substring "cr."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 GOODMAN

-0.76

 Zucker

-0.70

å¥

-0.68

 Mechdragon

-0.67

Materials

-0.65

ORGE

-0.65

EMENT

-0.64

 Werewolf

-0.64

 Dangerous

-0.64

 Malt

-0.63

POSITIVE LOGITS

ackers

0.97

cr

0.97

umbs

0.91

acker

0.90

adle

0.86

acies

0.86

usted

0.86

atell

0.84

omial

0.81

ursor

0.81

Activations Density 0.006%

terms related to physical destruction or violence

instances of the substring "cr."

No Comments

No Known Activations