INDEX

Explanations

phrases related to violence, particularly cyber violence and physical harm

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ellen

-0.91

izons

-0.83

çīĪ

-0.77

nir

-0.74

 Catalog

-0.73

NAS

-0.71

apter

-0.70

 Eternity

-0.68

 Startup

-0.68

velength

-0.68

POSITIVE LOGITS

 homicide

1.32

 homicides

1.28

 crimes

1.28

 crime

1.26

 violence

1.24

 perpetrator

1.22

 offenses

1.21

 violent

1.21

 rape

1.21

 rapes

1.17

Activations Density 0.451%

phrases related to violence, particularly cyber violence and physical harm

No Comments

No Known Activations