INDEX

Explanations

personalized messages indicating sharing or providing something to someone

oai_token-act-pair · gpt-3.5-turbo

references to the reader or audience

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ipal

-0.93

isa

-0.71

ended

-0.70

land

-0.66

peed

-0.65

ĸļ

-0.65

mare

-0.65

ela

-0.63

oldown

-0.63

aired

-0.63

POSITIVE LOGITS

 guys

1.76

 gentlemen

1.32

 ladies

1.16

 yourselves

1.14

 folks

1.13

tub

1.13

 boys

1.08

 dudes

1.08

 idiots

0.99

 readers

0.95

Activations Density 0.139%

personalized messages indicating sharing or providing something to someone

references to the reader or audience

No Comments

No Known Activations