INDEX

Explanations

abbreviations or terms related to a military context, such as "AU" and "NI."

oai_token-act-pair · gpt-3.5-turbo

mentions of the term "AU" and related currency or financial concepts

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

Redditor

-0.89

Ö¼

-0.88

stakes

-0.84

iffs

-0.81

dog

-0.80

idates

-0.78

idate

-0.76

tons

-0.76

ivas

-0.74

ows

-0.74

POSITIVE LOGITS

AU

1.06

AU

1.01

 hemor

0.74

RA

0.74

iliary

0.74

HAHA

0.73

 unemploy

0.73

 helic

0.73

 concess

0.72

RL

0.71

Activations Density 0.009%

abbreviations or terms related to a military context, such as "AU" and "NI."

mentions of the term "AU" and related currency or financial concepts

No Comments

No Known Activations