INDEX

Explanations

breaking news updates or headings

oai_token-act-pair · gpt-3.5-turbo

instances of the term "Breaking."

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

ogical

-0.76

inating

-0.73

eva

-0.71

haps

-0.71

omorphic

-0.70

istry

-0.70

wagen

-0.70

izing

-0.67

ateurs

-0.65

izers

-0.65

POSITIVE LOGITS

fast

0.99

 Breaking

0.81

bre

0.79

 Isles

0.78

Berry

0.77

AKING

0.77

neck

0.77

Bad

0.75

 Lives

0.73

ATH

0.71

Activations Density 0.018%

breaking news updates or headings

instances of the term "Breaking."

No Comments

No Known Activations