INDEX

Explanations

email addresses, particularly those associated with certain names and domains

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

 fitting

-0.77

jaw

-0.67

 apartheid

-0.65

©¶æ

-0.61

 validation

-0.59

 conformity

-0.59

 ancest

-0.58

CPR

-0.58

 Franch

-0.57

 totality

-0.56

POSITIVE LOGITS

gmail

1.17

yahoo

1.04

email

0.94

legram

0.84

mail

0.84

etsy

0.84

address

0.81

archives

0.79

dot

0.75

earth

0.73

Activations Density 0.024%

email addresses, particularly those associated with certain names and domains

No Comments

No Known Activations