INDEX

Explanations

email-related terms and references

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

 attractions

-0.91

 realism

-0.86

HG

-0.81

 stadiums

-0.73

 gears

-0.72

 progressing

-0.71

 Kingdoms

-0.69

 sights

-0.69

 industrialized

-0.68

 disadvantages

-0.68

POSITIVE LOGITS

mail

1.55

email

1.37

mails

1.31

 mail

1.27

message

1.25

 email

1.23

Mail

1.20

legram

1.15

letter

1.11

Email

1.11

Activations Density 0.033%

email-related terms and references

No Comments

No Known Activations