INDEX

Explanations

references to terrorism and extremist groups, particularly the Islamic State (ISIS)

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

eries

-0.75

yards

-0.74

 Seah

-0.74

 Winc

-0.69

ships

-0.69

 Constantin

-0.68

 Titanic

-0.68

WD

-0.67

 Vaugh

-0.67

é¾įå

-0.67

POSITIVE LOGITS

iyah

1.24

abad

1.18

istani

1.08

Ø±

1.02

 Hussein

1.00

 Sharif

0.99

 terrorists

0.99

Ø¯

0.96

 Sheikh

0.95

 cleric

0.94

Activations Density 4.251%

references to terrorism and extremist groups, particularly the Islamic State (ISIS)

No Comments

No Known Activations