INDEX

Explanations

phrases containing the term "minus"

oai_token-act-pair · gpt-3.5-turbo

references to negative values or errors in data reporting

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

lied

-0.78

uay

-0.76

istle

-0.73

urrent

-0.73

":[{"

-0.72

ollen

-0.72

stakes

-0.71

addock

-0.70

RAW

-0.70

omen

-0.69

POSITIVE LOGITS

 minus

1.11

minus

1.09

cules

0.95

(-

0.80

 insign

0.68

 fatig

0.68

mania

0.66

 rehe

0.66

 downside

0.64

 PLUS

0.64

Activations Density 0.007%

phrases containing the term "minus"

references to negative values or errors in data reporting

No Comments

No Known Activations