INDEX

Explanations

mentions or variations of the word "paper."

oai_token-act-pair · gpt-3.5-turbo

references to scholarly papers and articles

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 dism

-0.66

 disbelief

-0.64

Wow

-0.63

 status

-0.63

 condition

-0.61

 latitude

-0.60

 conditional

-0.60

 exclus

-0.60

 license

-0.60

 confirmed

-0.58

POSITIVE LOGITS

apers

4.55

aper

4.20

aping

1.78

aped

1.61

apes

1.51

ape

1.46

apest

1.34

acco

1.25

apor

1.24

aps

1.17

Activations Density 0.005%

mentions or variations of the word "paper."

references to scholarly papers and articles

No Comments

No Known Activations