INDEX

Explanations

phrases where someone is being labeled or called a negative term

oai_token-act-pair · gpt-3.5-turbo

instances of negative labels and accusations directed at individuals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

intent

-0.71

Instruct

-0.71

imates

-0.67

 adjoining

-0.65

ANS

-0.63

aday

-0.63

 appl

-0.63

grounds

-0.62

Edit

-0.62

MN

-0.62

POSITIVE LOGITS

 hoax

1.00

0.91

 liar

0.88

"'

0.82

 nuisance

0.80

 miracle

0.78

versive

0.77

 typo

0.77

 fraud

0.76

0.75

Activations Density 0.141%

phrases where someone is being labeled or called a negative term

instances of negative labels and accusations directed at individuals

No Comments

No Known Activations