INDEX

Explanations

verbs related to sharing or revealing information

oai_token-act-pair · gpt-3.5-turbo

terms related to revealing or withholding information and their consequences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

click

-0.83

Round

-0.81

raid

-0.75

mite

-0.73

 Catch

-0.72

ulate

-0.71

ENE

-0.71

flush

-0.70

urse

-0.70

inate

-0.69

POSITIVE LOGITS

 relinqu

1.62

 divul

1.61

 dissemin

1.60

 ascert

1.43

 assimil

1.43

 disav

1.41

 disreg

1.40

 solic

1.40

 obliter

1.39

 util

1.38

Activations Density 0.118%

verbs related to sharing or revealing information

terms related to revealing or withholding information and their consequences

No Comments

No Known Activations