INDEX

Explanations

mentions of evil in various contexts

oai_token-act-pair · gpt-3.5-turbo

instances and variations of the word "evil."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

raltar

-0.78

UNCH

-0.76

eding

-0.74

PsyNetMessage

-0.74

aro

-0.73

dropping

-0.73

kees

-0.71

akeru

-0.70

utra

-0.68

illation

-0.68

POSITIVE LOGITS

 incarn

1.00

 deeds

0.93

 deed

0.90

 twin

0.86

nesses

0.85

 genius

0.84

 mastermind

0.83

 lord

0.79

 evil

0.76

 empire

0.75

Activations Density 0.027%

mentions of evil in various contexts

instances and variations of the word "evil."

No Comments

No Known Activations