INDEX

Explanations

numeric values followed by a descriptive word or phrase like "roughly 9" or "roughly 10"

oai_token-act-pair · gpt-3.5-turbo

phrases that refer to approximate numerical values

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

oa

-0.76

iments

-0.73

 Bomber

-0.72

Reviewer

-0.70

ters

-0.67

ieu

-0.65

ysis

-0.64

orem

-0.63

rieg

-0.63

elsen

-0.63

POSITIVE LOGITS

Ĥİ

0.91

 ãĤµãĥ¼ãĥĨãĤ£ãĥ¯ãĥ³

0.78

Ń·

0.78

 approximate

0.77

 analogous

0.75

 sized

0.73

 contempor

0.70

 compr

0.69

atility

0.69

PsyNetMessage

0.69

Activations Density 0.009%

numeric values followed by a descriptive word or phrase like "roughly 9" or "roughly 10"

phrases that refer to approximate numerical values

No Comments

No Known Activations