INDEX

Explanations

names of individuals, possibly associated with news or controversy

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

mble

-0.76

undai

-0.56

urtles

-0.56

lehem

-0.55

eatures

-0.54

ometimes

-0.54

oenix

-0.54

ģĸ

-0.53

 streamed

-0.53

 brill

-0.51

POSITIVE LOGITS

's

1.26

ÃŃs

0.80

 personally

0.74

 himself

0.73

 Syndrome

0.71

existent

0.68

Â´

0.67

Care

0.67

tsy

0.66

Sr

0.65

Activations Density 2.623%

names of individuals, possibly associated with news or controversy

No Comments

No Known Activations