INDEX

Explanations

words related to traps or mechanisms designed to capture or deceive something

oai_token-act-pair · gpt-3.5-turbo

references to traps, both literal and metaphorical

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

edly

-0.87

issance

-0.84

sis

-0.82

verty

-0.70

played

-0.69

care

-0.69

sburg

-0.69

eda

-0.65

doms

-0.64

shire

-0.63

POSITIVE LOGITS

door

1.36

doors

1.22

idon

0.85

finding

0.81

 Hole

0.76

 traps

0.72

trap

0.69

 trap

0.67

hered

0.67

resses

0.67

Activations Density 0.037%

words related to traps or mechanisms designed to capture or deceive something

references to traps, both literal and metaphorical

No Comments

No Known Activations