INDEX

Explanations

introductions to lists or enumerations, particularly in the context of news articles, product releases, or company announcements.

oai_token-act-pair · claude-3-5-sonnet-20240620 Triggered by @aserino3

references to releases or publications

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

phrases related to new releases or announcements

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 UNCLASSIFIED

-0.95

lihood

-0.68

rador

-0.67

iferation

-0.65

apego

-0.65

ctory

-0.64

 Mubarak

-0.63

²¾

-0.62

ļéĨĴ

-0.62

Lv

-0.60

POSITIVE LOGITS

 cryptocurrency

0.66

 adorable

0.65

 indie

0.64

 Reddit

0.62

 frontman

0.62

 newest

0.61

 rapper

0.60

 hilar

0.60

 boldly

0.59

 downright

0.59

Activations Density 1.458%

introductions to lists or enumerations, particularly in the context of news articles, product releases, or company announcements.

references to releases or publications

phrases related to new releases or announcements

No Comments

No Known Activations