INDEX

Explanations

descriptions or mentions of physical appearances

oai_token-act-pair · gpt-3.5-turbo

references to physical appearance

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

tec

-0.75

stra

-0.73

ran

-0.70

rine

-0.70

eworks

-0.69

yards

-0.67

rap

-0.66

projects

-0.65

plex

-0.62

-0.61

POSITIVE LOGITS

 appearance

3.94

 Appearance

2.68

 appearances

2.65

 impression

1.63

earance

1.60

 appe

1.48

 appearing

1.40

Appearance

1.38

earances

1.34

 presentation

1.26

Activations Density 0.010%

descriptions or mentions of physical appearances

references to physical appearance

No Comments

No Known Activations