INDEX

Explanations

phrases related to classified information, security, and sensitive matters

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

·

-0.74

¶ħ

-0.68

\\\\\\\\

-0.68

kj

-0.64

ÅĤ

-0.64

here

-0.64

amus

-0.64

anon

-0.63

WW

-0.62

hip

-0.60

POSITIVE LOGITS

ised

1.11

izations

1.05

isations

1.01

ities

1.01

ties

0.98

 relativity

0.87

ariat

0.84

ty

0.83

isation

0.82

pared

0.82

Activations Density 4.806%

phrases related to classified information, security, and sensitive matters

No Comments

No Known Activations