INDEX

Explanations

instances where something is completely and thoroughly filled with something else

oai_token-act-pair · gpt-3.5-turbo

instances of the word "filled" 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

orem

-0.75

MIT

-0.72

arms

-0.71

orse

-0.69

arma

-0.69

hawks

-0.67

atives

-0.65

 attribution

-0.65

erity

-0.63

IDA

-0.63

POSITIVE LOGITS

 filled

3.41

filled

2.20

 filling

1.85

 fills

1.82

 fill

1.82

 packed

1.66

 populated

1.64

 overflowing

1.54

Fill

1.54

 littered

1.51

Activations Density 0.011%

instances where something is completely and thoroughly filled with something else

instances of the word "filled" in various contexts

No Comments

No Known Activations