INDEX

Explanations

phrases suggesting that the reader should check something out or try something

oai_token-act-pair · gpt-3.5-turbo

phrases encouraging action or engagement with content

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 foregoing

-0.73

 plurality

-0.64

 excess

-0.64

 dwelling

-0.63

roots

-0.62

 constant

-0.62

ordinate

-0.62

 terminating

-0.61

 dele

-0.58

 illnesses

-0.57

POSITIVE LOGITS

:)

1.06

 ðŁĻĤ

0.94

 firsthand

0.94

 yourself

0.91

 unfold

0.90

:-)

0.90

 HERE

0.90

;)

0.89

 aloud

0.89

0.88

Activations Density 0.179%

phrases suggesting that the reader should check something out or try something

phrases encouraging action or engagement with content

No Comments

No Known Activations