INDEX

Explanations

phrases related to security and surveillance

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

 Wikimedia

-0.77

Pg

-0.73

 Chau

-0.59

"}

-0.55

ãĥ¡

-0.55

malink

-0.54

 quot

-0.53

atre

-0.53

bourg

-0.53

gra

-0.53

POSITIVE LOGITS

DERR

0.90

how

0.85

why

0.83

 what

0.75

è¦ļéĨĴ

0.75

 beforehand

0.74

 whether

0.74

whether

0.73

why

0.72

that

0.70

Activations Density 2.691%

phrases related to security and surveillance

No Comments

No Known Activations