INDEX

Explanations

phrases related to acts of violence or conflict

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

ãĤ´ãĥ³

-0.99

 redes

-0.78

bet

-0.74

 spons

-0.73

soType

-0.72

HUD

-0.71

Ü

-0.70

cano

-0.68

aird

-0.67

algia

-0.65

POSITIVE LOGITS

 unsuspecting

1.31

 innocent

1.05

 civilians

1.00

 unarmed

0.95

 targets

0.93

 anyone

0.89

him

0.86

 opponents

0.85

 anybody

0.84

 somebody

0.83

Activations Density 1.695%

phrases related to acts of violence or conflict

No Comments

No Known Activations