INDEX

Explanations

dates indicated by 'nd'

oai_token-act-pair · gpt-3.5-turbo

instances of the ordinal indicator "nd" referring to dates or numerical sequences

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.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

 SPACE

-0.75

CHA

-0.69

BALL

-0.67

 portfolios

-0.66

flies

-0.66

IONS

-0.66

KEY

-0.64

ãĥĺ

-0.62

SHARE

-0.61

ÃįÃį

-0.61

POSITIVE LOGITS

ividual

1.12

nd

0.93

oyd

0.88

olin

0.88

amn

0.86

irty

0.85

ieth

0.82

cmp

0.80

oard

0.79

nen

0.78

Activations Density 0.009%

dates indicated by 'nd'

instances of the ordinal indicator "nd" referring to dates or numerical sequences

No Comments

No Known Activations