INDEX

Explanations

references to numbers and quantities

oai_token-act-pair · gpt-3.5-turbo

expressions of acknowledgment or recognition

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

ajo

-0.53

adelphia

-0.51

MSN

-0.50

unity

-0.50

ibrary

-0.50

ocene

-0.49

neys

-0.49

monton

-0.48

ippi

-0.47

and

-0.46

POSITIVE LOGITS

 consequently

0.92

 alike

0.89

 thereof

0.85

 consequ

0.81

 accordingly

0.80

 thereto

0.78

 thereafter

0.77

 therein

0.75

 thereby

0.71

SPONSORED

0.71

Activations Density 0.885%

references to numbers and quantities

expressions of acknowledgment or recognition

No Comments

No Known Activations