INDEX

Explanations

text surrounded by a lot of underscores

oai_token-act-pair · gpt-3.5-turbo

placeholder text or redacted information

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 divers

-0.90

oms

-0.70

ktop

-0.67

irection

-0.66

ifled

-0.64

 subord

-0.63

 Commission

-0.63

ena

-0.63

 division

-0.63

roads

-0.63

POSITIVE LOGITS

______

1.49

_______

1.39

_____

1.34

________________

1.28

___

1.28

________________________

1.25

________

1.24

____

1.24

________________________________________________________________

1.16

________________________________

1.05

Activations Density 0.013%

text surrounded by a lot of underscores

placeholder text or redacted information

No Comments

No Known Activations