INDEX

Explanations

the adjective "pure" followed by various different words

oai_token-act-pair · gpt-3.5-turbo

instances of the word "pure."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

è¦ļéĨĴ

-0.94

apter

-0.90

izoph

-0.87

 Peninsula

-0.82

otos

-0.77

 therap

-0.77

agements

-0.74

 WATCHED

-0.70

 Hilton

-0.69

assies

-0.69

POSITIVE LOGITS

waters

1.01

bred

0.90

blood

0.90

blooded

0.85

pure

0.81

ified

0.79

anus

0.76

water

0.75

 pure

0.74

enstein

0.72

Activations Density 0.016%

the adjective "pure" followed by various different words

instances of the word "pure."

No Comments

No Known Activations