INDEX

Explanations

the word "grave" with varying degrees of importance or relevance

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

ILA

-0.76

ivity

-0.70

BuyableInstoreAndOnline

-0.68

é¾

-0.66

Ack

-0.65

 Deng

-0.64

encer

-0.63

 CLIENT

-0.63

 warr

-0.61

ĩ

-0.60

POSITIVE LOGITS

lly

1.40

stones

0.93

 robbing

0.89

grave

0.86

lain

0.86

 grave

0.79

cover

0.79

lands

0.77

 robbers

0.76

grounds

0.76

Activations Density 0.016%

the word "grave" with varying degrees of importance or relevance

No Comments

No Known Activations