INDEX

Explanations

phrases related to official information or endorsements

oai_token-act-pair · gpt-3.5-turbo

references to testimonials or instances of validation

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 conscience

-0.62

tag

-0.61

 carn

-0.60

 peace

-0.60

 orche

-0.58

 Shack

-0.58

 torch

-0.58

 sense

-0.58

cap

-0.57

Tag

-0.56

POSITIVE LOGITS

ials

5.18

ial

2.58

ially

2.44

IAL

1.82

iations

1.72

iation

1.48

iating

1.42

iates

1.39

iated

1.38

iate

1.37

Activations Density 0.006%

phrases related to official information or endorsements

references to testimonials or instances of validation

No Comments

No Known Activations