INDEX

Explanations

phrases indicating instructions or promises to the reader

oai_token-act-pair · gpt-3.5-turbo

instances of the verb "be" in various forms and contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

arin

-0.70

 scrimmage

-0.64

 Implementation

-0.63

 Stain

-0.62

ottage

-0.61

iasco

-0.61

 truce

-0.61

Dialogue

-0.60

aceutical

-0.59

 guise

-0.59

POSITIVE LOGITS

 able

1.65

 amazed

1.27

 tempted

1.19

 surprised

1.15

 glad

1.14

 unable

1.05

 rewarded

1.05

 wondering

1.03

 pleasantly

1.02

 thankful

1.01

Activations Density 0.122%

phrases indicating instructions or promises to the reader

instances of the verb "be" in various forms and contexts

No Comments

No Known Activations