INDEX

Explanations

email-related phrases and terms

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

REL

-0.65

anova

-0.63

 ingred

-0.62

ahoo

-0.60

AVG

-0.60

caps

-0.59

 clicks

-0.57

uties

-0.57

wcs

-0.57

XL

-0.57

POSITIVE LOGITS

mail

0.84

 coli

0.84

vironment

0.82

mails

0.76

tainment

0.74

ionage

0.70

theless

0.68

cigarette

0.67

Reloaded

0.66

hardt

0.66

Activations Density 1.097%

email-related phrases and terms

No Comments

No Known Activations