INDEX

Explanations

words related to the end or conclusion

oai_token-act-pair · gpt-3.5-turbo

instances of "end."

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

CHA

-0.71

bryce

-0.67

«ĺ

-0.67

ppo

-0.65

æ©Ł

-0.64

terness

-0.64

IGHTS

-0.64

Ĥİ

-0.63

 subur

-0.63

KING

-0.62

POSITIVE LOGITS

end

1.14

angered

1.01

ends

0.93

ending

0.91

lich

0.91

ocrin

0.89

ended

0.85

urance

0.82

orph

0.81

ocrine

0.78

Activations Density 0.013%

words related to the end or conclusion

instances of "end."

No Comments

No Known Activations