INDEX

Explanations

usernames or aliases, likely related to online communities or platforms

oai_token-act-pair · gpt-3.5-turbo

occurrences of non-word characters and punctuation indicators

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 destro

-0.65

rul

-0.54

 livest

-0.53

 Vaugh

-0.50

 convol

-0.49

 advoc

-0.47

 helicop

-0.47

 predec

-0.47

 distingu

-0.46

',"

-0.45

POSITIVE LOGITS

Cub

0.56

º

0.53

SCP

0.52

CVE

0.49

 largeDownload

0.49

Vers

0.49

Ī

0.45

ËĪ

0.45

UTERS

0.44

 âĢº

0.43

Activations Density 0.531%

usernames or aliases, likely related to online communities or platforms

occurrences of non-word characters and punctuation indicators

No Comments

No Known Activations