INDEX

Explanations

the keyword "AV" followed by a number, potentially related to a technical or specialized term

oai_token-act-pair · gpt-3.5-turbo

mentions of audio-visual content or media-related terms

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

 rift

-0.67

 Jere

-0.65

 Raleigh

-0.65

 Franklin

-0.63

 Phil

-0.63

Neh

-0.62

 reflection

-0.61

 grit

-0.61

 depression

-0.61

grave

-0.60

POSITIVE LOGITS

AV

4.18

AV

1.88

av

1.84

Av

1.61

AW

1.55

AX

1.44

CV

1.39

AK

1.32

AF

1.31

UV

1.29

Activations Density 0.009%

the keyword "AV" followed by a number, potentially related to a technical or specialized term

mentions of audio-visual content or media-related terms

No Comments

No Known Activations