INDEX

Explanations

references to a character named "Nova 9."

oai_token-act-pair · gpt-3.5-turbo

mentions of "Nova"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

—-

-0.71

 oblig

-0.68

alks

-0.66

 wing

-0.64

anks

-0.63

 ridden

-0.63

li

-0.62

gged

-0.62

ALK

-0.57

PF

-0.57

POSITIVE LOGITS

 Nova

4.23

 Halifax

1.85

nova

1.57

Nun

1.56

 Newfoundland

1.53

ova

1.42

 Manitoba

1.25

 Saskatchewan

1.25

 Scotia

1.25

 Quebec

1.22

Activations Density 0.014%

references to a character named "Nova 9."

mentions of "Nova"

No Comments

No Known Activations