INDEX

Explanations

phrases indicating that something is morally incorrect or unacceptable

oai_token-act-pair · gpt-3.5-turbo

phrases indicating the absence of wrongdoing

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

incinn

-0.82

cit

-0.81

soType

-0.77

xit

-0.70

earchers

-0.69

weeney

-0.68

ulum

-0.66

wit

-0.66

cum

-0.65

Ri

-0.64

POSITIVE LOGITS

headed

0.75

mouth

0.74

 wrong

0.70

 wing

0.69

eous

0.68

 havoc

0.67

 behaviour

0.65

nered

0.63

 flank

0.63

doing

0.62

Activations Density 0.011%

phrases indicating that something is morally incorrect or unacceptable

phrases indicating the absence of wrongdoing

No Comments

No Known Activations