INDEX

Explanations

years from the 80s

oai_token-act-pair · gpt-3.5-turbo

numerical values, particularly those associated with percentages and years

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

zag

-0.70

actor

-0.66

uden

-0.65

peat

-0.63

assic

-0.61

 illusion

-0.61

plom

-0.60

Lens

-0.60

akers

-0.60

ussia

-0.59

POSITIVE LOGITS

1.07

0.91

0.83

dayName

0.78

itement

0.77

idently

0.77

0.76

0.75

itability

0.75

Activations Density 0.056%

years from the 80s

numerical values, particularly those associated with percentages and years

No Comments

No Known Activations