INDEX

Explanations

numerical identifiers that refer to user profile numbers and comments in a forum-like setting

oai_token-act-pair · gpt-3.5-turbo

numeric identifiers related to user posts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

 psychiat

-0.76

 tabl

-0.73

 cinema

-0.73

ĸļå£«

-0.70

numbered

-0.69

 miscar

-0.68

¥ŀ

-0.67

 estranged

-0.66

 institutions

-0.66

ãĥķãĤ¡

-0.65

POSITIVE LOGITS

 Quote

1.06

 Alright

0.88

Hmm

0.87

 Seems

0.87

 Interesting

0.87

Wow

0.85

 Congratulations

0.84

 Awesome

0.83

 Thanks

0.83

EDIT

0.79

Activations Density 0.023%

numerical identifiers that refer to user profile numbers and comments in a forum-like setting

numeric identifiers related to user posts

No Comments

No Known Activations