INDEX

Explanations

the word "seen" followed by a high number

oai_token-act-pair · gpt-3.5-turbo

instances of the word "seen" 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 @ 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

rawdownloadcloneembedreportprint

-0.73

verning

-0.60

 Pledge

-0.59

bing

-0.59

 guilt

-0.56

 moratorium

-0.56

FAQ

-0.56

mination

-0.55

forward

-0.55

wald

-0.55

POSITIVE LOGITS

 seen

3.46

 Seen

2.41

seen

2.16

 witnessed

1.79

 viewed

1.79

 heard

1.78

 shown

1.74

 spotted

1.55

 pictured

1.53

 encountered

1.45

Activations Density 0.025%

the word "seen" followed by a high number

instances of the word "seen" in various contexts

No Comments

No Known Activations