INDEX

Explanations

assertive statements, usually related to serious matters

oai_token-act-pair · gpt-3.5-turbo

instances of the word "serious" and related terms

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

 epit

-0.65

oS

-0.64

 partition

-0.64

 clock

-0.63

via

-0.62

azon

-0.62

 hereafter

-0.62

 feed

-0.62

 freely

-0.61

Ft

-0.60

POSITIVE LOGITS

serious

3.71

 Serious

2.50

 seriousness

2.33

 serious

2.23

 seriously

1.32

grave

1.18

severe

1.16

 Seriously

1.11

threatening

1.08

 severe

1.06

Activations Density 0.023%

assertive statements, usually related to serious matters

instances of the word "serious" and related terms

No Comments

No Known Activations