INDEX

Explanations

print-related words or phrases

oai_token-act-pair · gpt-3.5-turbo

references to printed materials or print media

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

 basin

-0.66

oxy

-0.65

itability

-0.64

æĢ

-0.63

 domin

-0.62

icably

-0.62

kh

-0.62

 stranger

-0.60

ãĥ¤

-0.60

 nowhere

-0.59

POSITIVE LOGITS

 Print

3.99

Print

3.15

 print

2.40

 prints

2.21

print

2.21

 Printing

2.12

 Printed

1.92

prints

1.85

 printing

1.81

printed

1.69

Activations Density 0.012%

print-related words or phrases

references to printed materials or print media

No Comments

No Known Activations