INDEX

Explanations

phrases related to negative actions or behaviors

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

edia

-0.78

_>

-0.76

itan

-0.74

akeru

-0.72

ocobo

-0.71

aeda

-0.71

udeau

-0.69

Downloadha

-0.65

 Roosevelt

-0.65

btn

-0.65

POSITIVE LOGITS

cery

1.00

 smelling

0.89

sie

0.84

terness

0.82

 foul

0.78

mouth

0.76

0.75

nesses

0.73

eners

0.70

rance

0.70

Activations Density 0.018%

phrases related to negative actions or behaviors

No Comments

No Known Activations