INDEX

Explanations

proper nouns, specifically names of people, likely focusing on multiple occurrences of the same name

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

cffffcc

-1.01

NetMessage

-0.88

ĺħ

-0.77

phies

-0.75

vana

-0.74

Reply

-0.73

ttes

-0.73

inventoryQuantity

-0.72

sites

-0.71

tnc

-0.70

POSITIVE LOGITS

 Bowie

1.06

 Cameron

0.98

 Axel

0.94

 Beckham

0.90

son

0.90

 Petraeus

0.89

 Rockefeller

0.87

Ic

0.86

 Barton

0.84

 Foster

0.82

Activations Density 0.044%

proper nouns, specifically names of people, likely focusing on multiple occurrences of the same name

No Comments

No Known Activations