INDEX

Explanations

references to evil or negativity

oai_token-act-pair · gpt-3.5-turbo

references to the concept of "evil."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ribe

-0.75

UNCH

-0.74

GN

-0.71

Lago

-0.70

ainers

-0.70

PsyNetMessage

-0.70

ribes

-0.70

cially

-0.69

RESULTS

-0.69

aro

-0.67

POSITIVE LOGITS

 incarn

1.01

 deed

0.89

 undermin

0.89

 deeds

0.89

 genius

0.87

 evil

0.86

 mastermind

0.86

nesses

0.83

 lord

0.83

 twin

0.83

Activations Density 0.015%

references to evil or negativity

references to the concept of "evil."

No Comments

No Known Activations