INDEX

Explanations

commands or suggestions to forget something or specific actions

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

CVE

-0.82

Led

-0.79

PokÃ©

-0.74

dn

-0.70

can

-0.70

Redditor

-0.69

QUI

-0.67

amen

-0.67

MpServer

-0.66

FLAG

-0.66

POSITIVE LOGITS

 altogether

0.97

 precon

0.96

 inhib

0.87

 illusions

0.86

 pesky

0.86

 reliance

0.84

 superflu

0.83

 entirely

0.83

 notions

0.83

 distractions

0.81

Activations Density 0.152%

commands or suggestions to forget something or specific actions

No Comments

No Known Activations