INDEX

Explanations

phrases related to criminal activities and theft

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

ratulations

-0.88

merce

-0.87

retty

-0.84

ICA

-0.78

video

-0.78

SPONSORED

-0.77

ilater

-0.73

TEXT

-0.70

QUI

-0.69

meric

-0.69

POSITIVE LOGITS

 afar

0.99

 whence

0.98

 thence

0.90

 unsuspecting

0.88

 underneath

0.88

 beneath

0.86

 Palest

0.84

 circulation

0.81

 somewhere

0.74

 abroad

0.72

Activations Density 0.100%

phrases related to criminal activities and theft

No Comments

No Known Activations