INDEX

Explanations

dialogues including quotation marks

oai_token-act-pair · gpt-3.5-turbo

direct speech or quotes in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 characterized

-0.74

 conclud

-0.73

 upstream

-0.72

 downstream

-0.72

 valued

-0.71

 frontline

-0.70

 footing

-0.70

 ticket

-0.69

¥ŀ

-0.69

 pole

-0.68

POSITIVE LOGITS

Oh

1.18

Yeah

1.16

wcsstore

1.11

Yes

1.07

Fuck

1.06

Hmm

1.04

YES

1.04

Sorry

1.02

Huh

1.02

Damn

1.02

Activations Density 0.094%

dialogues including quotation marks

direct speech or quotes in the text

No Comments

No Known Activations