INDEX

Explanations

words or phrases that resemble computer code, such as variable names and function declarations

oai_token-act-pair · gpt-3.5-turbo

identifiers or keys associated with entities or objects

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 Melania

-0.90

 stocking

-0.80

asta

-0.76

 Anne

-0.75

ega

-0.74

 Melanie

-0.73

 Martin

-0.72

andel

-0.72

 adulthood

-0.72

 Billy

-0.70

POSITIVE LOGITS

1.84

1.72

QU

1.50

1.49

qu

1.46

qq

1.42

qs

1.42

QL

1.42

qi

1.41

Activations Density 0.210%

words or phrases that resemble computer code, such as variable names and function declarations

identifiers or keys associated with entities or objects

No Comments

No Known Activations