INDEX

Explanations

phrases related to job roles and responsibilities

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate restrictions or limitations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Miche

-0.59

Redditor

-0.58

bath

-0.56

RF

-0.55

Stars

-0.55

ogenesis

-0.55

 Outbreak

-0.52

ty

-0.51

TX

-0.51

 Examples

-0.51

POSITIVE LOGITS

 previously

1.61

 hitherto

1.36

 formerly

1.28

 originally

1.23

 hoped

1.20

 anticipated

1.13

 intended

1.05

 planned

1.03

 normally

1.02

planned

1.02

Activations Density 0.761%

phrases related to job roles and responsibilities

phrases that indicate restrictions or limitations

No Comments

No Known Activations