INDEX

Explanations

proper nouns and specific terms related to politics and opinions

oai_token-act-pair · gpt-3.5-turbo

prominent names and figures in a political or controversial context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

ciplinary

-0.82

 facility

-0.78

 preliminary

-0.78

 recreation

-0.75

 vacant

-0.75

 Alpine

-0.75

ĸļ

-0.74

 rehabilitation

-0.71

 quarry

-0.70

 aerial

-0.70

POSITIVE LOGITS

ghazi

1.05

OTUS

1.02

addafi

0.97

Obama

0.95

Hillary

0.87

Gamer

0.87

 Miliband

0.85

ISM

0.85

ocrates

0.85

Reddit

0.84

Activations Density 0.420%

proper nouns and specific terms related to politics and opinions

prominent names and figures in a political or controversial context

No Comments

No Known Activations