INDEX

Explanations

dates written as "Feb" followed by a numerical day of the month

oai_token-act-pair · gpt-3.5-turbo

occurrences of specific dates

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

ographies

-0.80

advertising

-0.67

ross

-0.67

ography

-0.67

ographical

-0.63

alogue

-0.61

respective

-0.60

 comprehens

-0.59

 invested

-0.59

itect

-0.58

POSITIVE LOGITS

Feb

3.32

Feb

2.57

Jan

2.26

Aug

2.24

Nov

2.18

Oct

2.15

Apr

2.13

 Sept

2.02

 February

1.99

Dec

1.97

Activations Density 0.005%

dates written as "Feb" followed by a numerical day of the month

occurrences of specific dates

No Comments

No Known Activations