INDEX

Explanations

phrases related to controversial statements or actions made by public figures

oai_token-act-pair · gpt-3.5-turbo

instances of names and titles associated with accusations or negative labels

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

icion

-0.80

enture

-0.69

itely

-0.68

regate

-0.68

ovember

-0.63

pta

-0.62

few

-0.61

ptoms

-0.61

éĥ

-0.61

gomery

-0.61

POSITIVE LOGITS

 unacceptable

1.08

 unreliable

0.99

 irresponsible

0.99

 unfit

0.98

 unsu

0.97

 obsolete

0.97

 unworthy

0.95

 unethical

0.95

 illegitimate

0.92

"'

0.91

Activations Density 0.141%

phrases related to controversial statements or actions made by public figures

instances of names and titles associated with accusations or negative labels

No Comments

No Known Activations