INDEX

Explanations

the word 'next'

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

 Feldman

-0.78

ocker

-0.74

ordes

-0.68

enance

-0.67

lees

-0.66

baugh

-0.66

 Ãĸ

-0.65

ans

-0.63

lee

-0.63

cks

-0.62

POSITIVE LOGITS

 week

0.98

 door

0.91

 month

0.90

door

0.89

 generation

0.86

ĻĤ

0.83

 year

0.83

 semester

0.82

 generations

0.81

STEP

0.79

Activations Density 0.049%

the word 'next'

No Comments

No Known Activations

the word 'next'

No Comments

No Known Activations