INDEX

Explanations

numeric ranks or ratings

oai_token-act-pair · gpt-3.5-turbo

statements about rankings and positions in various contexts

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

 fold

-0.64

unity

-0.61

cms

-0.60

 refinery

-0.59

ABE

-0.55

Ħ¢

-0.55

credit

-0.54

 umbrella

-0.53

dy

-0.53

 withdrawals

-0.52

POSITIVE LOGITS

 ninth

1.11

 seventh

1.11

 sixth

1.11

 eighth

1.09

teenth

1.08

 among

1.07

 fifth

1.03

 tenth

0.99

 fourth

0.98

 amongst

0.97

Activations Density 0.075%

numeric ranks or ratings

statements about rankings and positions in various contexts

No Comments

No Known Activations