INDEX

Explanations

instances of the word "lost"

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

sis

-0.69

ansky

-0.65

 repr

-0.65

ECK

-0.64

osi

-0.63

 barr

-0.63

 easing

-0.63

zer

-0.61

 prepared

-0.60

instein

-0.60

POSITIVE LOGITS

 souls

0.94

 sight

0.87

lust

0.85

 Souls

0.83

luster

0.81

erness

0.81

hearted

0.79

 innocence

0.78

iscover

0.76

 forever

0.76

Activations Density 0.022%

instances of the word "lost"

No Comments

No Known Activations