INDEX

Explanations

phrases related to news articles and reports

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

seys

-0.76

estern

-0.68

ernels

-0.65

aiman

-0.65

mble

-0.65

STD

-0.65

cgi

-0.65

kees

-0.63

interrupted

-0.62

thirst

-0.62

POSITIVE LOGITS

 foremost

0.95

imester

0.95

 bidder

0.75

eve

0.74

highest

0.71

century

0.70

generation

0.68

 gasp

0.68

ever

0.66

minster

0.66

Activations Density 1.093%

phrases related to news articles and reports

No Comments

No Known Activations