INDEX

Explanations

phrases related to actions involving authority figures or legal matters

oai_token-act-pair · gpt-3.5-turbo

phrases related to sources and citations in a document

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

ptoms

-0.79

endars

-0.71

isphere

-0.70

 responses

-0.70

 havens

-0.68

artifacts

-0.68

 endeavors

-0.67

Cause

-0.67

 trillions

-0.66

 sexes

-0.66

POSITIVE LOGITS

 hers

0.91

 theirs

0.88

 another

0.76

lier

0.76

 Uzbek

0.75

 yours

0.68

 Mahmoud

0.67

 someone

0.67

ãĢĲ

0.66

 Yugoslavia

0.66

Activations Density 0.554%

phrases related to actions involving authority figures or legal matters

phrases related to sources and citations in a document

No Comments

No Known Activations