INDEX

Explanations

words related to law enforcement and crime reporting

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

erella

-0.77

ãĥ¯ãĥ³

-0.68

ificent

-0.64

 earthqu

-0.64

nox

-0.64

risome

-0.64

PDATE

-0.63

staking

-0.63

vell

-0.63

realDonaldTrump

-0.62

POSITIVE LOGITS

 syndrome

1.01

(<

0.92

vironments

0.89

(>

0.86

 Syndrome

0.85

ancies

0.77

(.

0.73

hips

0.72

ologies

0.72

(%)

0.71

Activations Density 3.150%

words related to law enforcement and crime reporting

No Comments

No Known Activations