INDEX

Explanations

violent or aggressive actions and incidents

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

QL

-0.82

ologies

-0.77

PRESS

-0.73

PN

-0.70

UFF

-0.68

IGN

-0.68

IPP

-0.68

ItemThumbnailImage

-0.67

lance

-0.67

HCR

-0.66

POSITIVE LOGITS

 bushes

1.20

 doorway

1.16

 ceiling

1.02

 shoulder

1.00

 windshield

0.99

 shoulders

0.99

 vents

0.97

 turret

0.96

 ledge

0.95

 cliffs

0.95

Activations Density 5.258%

violent or aggressive actions and incidents

No Comments

No Known Activations