INDEX

Explanations

phrases or keywords related to being reported or identified for scrutiny or caution

oai_token-act-pair · gpt-3.5-turbo

instances of the word "flagged."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

elf

-0.77

enny

-0.72

oi

-0.72

pite

-0.72

oral

-0.71

ppard

-0.69

anny

-0.69

ofi

-0.69

uper

-0.68

aaaa

-0.68

POSITIVE LOGITS

 flagged

1.43

staff

0.70

pole

0.69

lights

0.66

 flags

0.65

listed

0.62

ging

0.62

Benz

0.62

ãĥĨ

0.61

 headlights

0.61

Activations Density 0.003%

phrases or keywords related to being reported or identified for scrutiny or caution

instances of the word "flagged."

No Comments

No Known Activations