INDEX

Explanations

references to specific messaging platforms, especially WhatsApp

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

Negative Logits

©¶æ

-0.87

herty

-0.79

abama

-0.78

gling

-0.76

ãĤ½

-0.74

 Tolkien

-0.70

arson

-0.69

opez

-0.69

arde

-0.68

bage

-0.68

POSITIVE LOGITS

 Messenger

0.95

 subscriptions

0.92

 Users

0.91

 users

0.89

Chat

0.87

 messaging

0.86

 Messages

0.84

 Chat

0.82

 chats

0.82

 subscribers

0.82

Activations Density 0.048%

references to specific messaging platforms, especially WhatsApp

No Known Activations