INDEX

Explanations

phrases related to crimes and investigations, specifically emphasizing immediate action in response to criminal 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_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

NetMessage

-0.73

escription

-0.71

ttes

-0.68

 Sack

-0.67

Downloadha

-0.66

 lodge

-0.65

Rav

-0.64

 Rost

-0.63

ĳå£«

-0.63

 Morales

-0.62

POSITIVE LOGITS

itating

1.20

itates

1.18

itations

1.16

balanced

1.15

mer

1.13

itated

1.12

manent

1.05

ply

1.01

mediate

1.01

bal

1.00

Activations Density 0.012%

phrases related to crimes and investigations, specifically emphasizing immediate action in response to criminal incidents

No Comments

No Known Activations