INDEX

Explanations

phrases related to contrasting statements or pieces of information

oai_token-act-pair · gpt-3.5-turbo

the word "and" used in a variety of contexts to connect phrases or ideas

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

ses

-0.70

Fine

-0.70

Limited

-0.67

":-

-0.67

Glass

-0.66

Hide

-0.64

AU

-0.64

Ur

-0.63

asper

-0.63

kell

-0.63

POSITIVE LOGITS

romeda

0.91

 consequently

0.90

 preferably

0.89

 possibly

0.88

 hopefully

0.87

 vice

0.87

 therefore

0.87

 rightly

0.86

 optionally

0.86

 rightfully

0.84

Activations Density 0.083%

phrases related to contrasting statements or pieces of information

the word "and" used in a variety of contexts to connect phrases or ideas

No Comments

No Known Activations