INDEX

Explanations

the sequence "hom" and related words

oai_token-act-pair · gpt-3.5-turbo

references to different hominid species and their characteristics

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

cia

-0.80

 Browne

-0.78

vision

-0.76

 Notification

-0.75

bolt

-0.71

toggle

-0.68

 Unlock

-0.67

UGC

-0.66

OWER

-0.66

 Rolls

-0.66

POSITIVE LOGITS

hom

3.82

hom

1.95

Hom

1.92

 heter

1.68

Hom

1.65

 Homo

1.28

 homophobic

1.15

heter

1.06

mon

1.02

 home

0.97

Activations Density 0.010%

the sequence "hom" and related words

references to different hominid species and their characteristics

No Comments

No Known Activations