INDEX

Explanations

politically related information

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

 ......

-0.25

 [...]

-0.25

 Sith

-0.24

 Zerg

-0.23

ICO

-0.23

 StarCraft

-0.23

 crochet

-0.23

 MPEG

-0.23

 Minecraft

-0.23

 amazing

-0.22

POSITIVE LOGITS

etheless

0.32

warts

0.31

abal

0.30

afer

0.30

ibles

0.29

ħĭ

0.29

²¾

0.29

tains

0.29

ignty

0.28

issance

0.28

Activations Density 10.513%

politically related information

No Comments

No Known Activations