INDEX

Explanations

parts of the text with the word "pred" followed by a numerical value, such as "pred10."

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

ASE

-0.69

BOX

-0.68

Ø©

-0.65

 Scotia

-0.65

MENT

-0.64

bians

-0.64

glers

-0.63

 twist

-0.62

Hub

-0.61

UGE

-0.61

POSITIVE LOGITS

efined

1.27

nis

1.11

acent

1.04

essor

1.02

icip

0.97

icated

0.96

awn

0.95

ents

0.94

icates

0.94

etermin

0.94

Activations Density 0.064%

parts of the text with the word "pred" followed by a numerical value, such as "pred10."

No Comments

No Known Activations