INDEX

Explanations

words related to political figures and events

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

IUM

-0.80

lished

-0.58

 pressures

-0.57

¶ħ

-0.57

 popcorn

-0.56

 hazards

-0.55

 surface

-0.55

 calibrated

-0.55

 Downs

-0.54

Ĥİ

-0.54

POSITIVE LOGITS

Allah

0.82

wife

0.80

hood

0.78

daughter

0.77

abet

0.73

brother

0.73

backer

0.71

chief

0.71

aiden

0.70

son

0.70

Activations Density 0.069%

words related to political figures and events

No Comments

No Known Activations