INDEX

Explanations

days of the week - specifically, the neuron is likely finding mentions of specific days in news articles or documents

oai_token-act-pair · gpt-3.5-turbo

days of the week or specific dates mentioned in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

atorial

-0.71

ced

-0.67

tumblr

-0.66

cius

-0.66

Vin

-0.65

gy

-0.64

-0.63

tnc

-0.62

wb

-0.61

mental

-0.61

POSITIVE LOGITS

 afternoon

1.05

 morning

1.00

 evening

0.94

 night

0.83

 redund

0.79

 mornings

0.73

 revealing

0.71

 emphatically

0.67

»Ĵ

0.66

 Tonight

0.65

Activations Density 0.112%

days of the week - specifically, the neuron is likely finding mentions of specific days in news articles or documents

days of the week or specific dates mentioned in the text

No Comments

No Known Activations