INDEX

Explanations

words related to threats or potential harm

oai_token-act-pair · gpt-3.5-turbo

references to the concept of danger

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

olitan

-0.90

ergy

-0.80

pel

-0.73

elf

-0.73

states

-0.72

orney

-0.72

venants

-0.71

eenth

-0.70

essee

-0.69

owned

-0.69

POSITIVE LOGITS

 Danger

0.97

 endanger

0.92

ously

0.89

 lurking

0.88

lur

0.86

 danger

0.85

danger

0.82

lessly

0.80

crow

0.76

 undermin

0.75

Activations Density 0.025%

words related to threats or potential harm

references to the concept of danger

No Comments

No Known Activations