INDEX

Explanations

names of people

oai_token-act-pair · gpt-3.5-turbo

proper nouns and capitalized identifiers

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 referen

-0.81

 contempor

-0.77

 metic

-0.75

 bullet

-0.75

cow

-0.74

 bean

-0.73

 confer

-0.73

 delegate

-0.71

 hemor

-0.70

 finger

-0.69

POSITIVE LOGITS

ses

1.13

asses

1.12

les

1.12

ps

1.12

bles

1.11

als

1.11

ents

1.10

zes

1.09

ds

1.07

ues

1.07

Activations Density 0.154%

names of people

proper nouns and capitalized identifiers

No Comments

No Known Activations