INDEX

Explanations

words related to the concept of harm or potential harm

oai_token-act-pair · gpt-3.5-turbo

instances of the word "harm" and its context within discussions of its effects

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 liner

-0.74

zsche

-0.69

ARCH

-0.68

KER

-0.66

ulating

-0.62

ulous

-0.61

ularity

-0.61

ulates

-0.60

ULAR

-0.60

olitan

-0.59

POSITIVE LOGITS

onies

1.37

lessly

1.25

 harm

0.98

 inflicted

0.93

 harms

0.91

fully

0.90

lessness

0.87

ful

0.85

benefit

0.83

fulness

0.82

Activations Density 0.022%

words related to the concept of harm or potential harm

instances of the word "harm" and its context within discussions of its effects

No Comments

No Known Activations