INDEX

Explanations

mentions related to the television network NBC

oai_token-act-pair · gpt-3.5-turbo

instances of the acronym "NBC"

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

aunder

-0.73

ebook

-0.72

andr

-0.71

erva

-0.67

adding

-0.66

piration

-0.66

arians

-0.65

============

-0.65

rosse

-0.64

 McCartney

-0.64

POSITIVE LOGITS

Universal

1.19

NBC

0.91

NBC

0.89

SN

0.76

 News

0.73

estern

0.72

 affiliate

0.72

WS

0.71

 Interstellar

0.69

 Universal

0.68

Activations Density 0.008%

mentions related to the television network NBC

instances of the acronym "NBC"

No Comments

No Known Activations