INDEX

Explanations

references to personal information or instructions specifically directed at recipients

oai_token-act-pair · gpt-3.5-turbo

phrases that emphasize the word "your" in relation to inquiries, account information, or personal details

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

forth

-0.76

verb

-0.74

 Canaver

-0.72

 Secondly

-0.70

 boils

-0.65

 Goes

-0.64

 airs

-0.63

 discipl

-0.63

atter

-0.63

 alike

-0.62

POSITIVE LOGITS

own

1.24

 favorite

1.21

 favourite

1.20

 preferred

1.02

 favorites

0.98

 desired

0.98

 browser

0.93

 favourites

0.93

 choice

0.90

 chosen

0.90

Activations Density 0.122%

references to personal information or instructions specifically directed at recipients

phrases that emphasize the word "your" in relation to inquiries, account information, or personal details

No Comments

No Known Activations