INDEX

Explanations

words related to anti-sentiments, such as "anti-gun," "anti-abortion," and "anti-Christian."

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_fs12288-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

12,288

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

ĸļ

-0.74

 Sins

-0.69

ãĤ¼ãĤ¦ãĤ¹

-0.68

 Pelicans

-0.65

 resid

-0.64

ODY

-0.63

 grapes

-0.62

 Halls

-0.62

 Tycoon

-0.62

 Gorge

-0.61

POSITIVE LOGITS

abortion

1.22

establishment

1.16

war

1.12

hero

0.98

strate

0.91

immigrant

0.89

Semitic

0.89

-'

0.89

overty

0.89

inflammatory

0.87

Activations Density 0.042%

words related to anti-sentiments, such as "anti-gun," "anti-abortion," and "anti-Christian."

No Comments

No Known Activations