INDEX

Explanations

phrases related to related content or topics

oai_token-act-pair · gpt-3.5-turbo

multiple instances of the word "Related" likely indicating a focus on connections or associations between different topics

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 swords

-0.66

 breath

-0.65

 mana

-0.65

 being

-0.63

 arrived

-0.62

 colours

-0.61

 humour

-0.61

 liber

-0.60

oct

-0.60

 cherry

-0.60

POSITIVE LOGITS

Related

3.81

 Related

2.08

RELATED

2.08

related

1.65

SPONSORED

1.46

Specific

1.39

Recommended

1.37

Associated

1.34

MORE

1.32

Gallery

1.31

Activations Density 0.016%

phrases related to related content or topics

multiple instances of the word "Related" likely indicating a focus on connections or associations between different topics

No Comments

No Known Activations