INDEX

Explanations

phrases related to errors and mistakes

oai_token-act-pair · gpt-3.5-turbo

mentions of errors and issues with performance or accuracy

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

electric

-0.78

tsky

-0.78

apeake

-0.77

apy

-0.76

amen

-0.75

nai

-0.75

estine

-0.74

edom

-0.73

bledon

-0.72

rons

-0.70

POSITIVE LOGITS

ously

0.85

gered

0.83

 guiActiveUn

0.82

 margin

0.79

uracy

0.78

 error

0.75

 deceive

0.71

 prone

0.70

 dece

0.69

 errors

0.68

Activations Density 0.022%

phrases related to errors and mistakes

mentions of errors and issues with performance or accuracy

No Comments

No Known Activations