INDEX

Explanations

positive attributes related to ethics and morality

oai_token-act-pair · gpt-3.5-turbo

references to compassion and greed

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Marble

-0.88

 Surface

-0.80

 Signal

-0.73

 Completed

-0.72

 Offline

-0.71

Rou

-0.70

Ja

-0.70

 Brotherhood

-0.69

SIG

-0.68

mol

-0.67

POSITIVE LOGITS

 compassion

2.65

 greed

2.56

 compassionate

1.96

 greedy

1.59

 selfish

1.47

 altru

1.42

 disdain

1.19

 condesc

1.11

 generosity

1.09

 bene

1.09

Activations Density 0.022%

positive attributes related to ethics and morality

references to compassion and greed

No Comments

No Known Activations