INDEX

Explanations

words related to safety

oai_token-act-pair · gpt-3.5-turbo

references to safety

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

utenberg

-0.77

é¾

-0.77

itus

-0.77

ovi

-0.70

ities

-0.69

iets

-0.68

isson

-0.68

rum

-0.67

gae

-0.67

hawks

-0.63

POSITIVE LOGITS

 safe

3.85

safe

2.87

 safer

2.64

 Safe

2.52

 safest

2.33

Safe

2.18

 safely

2.10

 unsafe

2.10

 safety

1.74

SAF

1.69

Activations Density 0.011%

words related to safety

references to safety

No Comments

No Known Activations