INDEX

Explanations

the word "tape."

oai_token-act-pair · gpt-3.5-turbo

references to tape or recordings

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

 Democr

-0.71

erate

-0.70

ablishment

-0.70

ationally

-0.69

udeb

-0.65

abad

-0.65

 Hague

-0.65

finder

-0.63

 âĺ

-0.62

HCR

-0.61

POSITIVE LOGITS

 tape

1.38

 Tape

1.23

 tapes

1.15

 cassette

0.93

 recorder

0.92

 taped

0.89

clip

0.83

entimes

0.80

 recordings

0.76

 conclud

0.75

Activations Density 0.007%

the word "tape."

references to tape or recordings

No Comments

No Known Activations