INDEX

Explanations

phrases related to legal terms and violations

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate conditional scenarios or alternatives

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

 Pony

-0.83

ires

-0.79

ocracy

-0.73

ETS

-0.69

emi

-0.66

onday

-0.65

idem

-0.63

 Outbreak

-0.61

IDENT

-0.61

ocrat

-0.60

POSITIVE LOGITS

acle

1.27

nam

1.21

acles

1.20

chard

1.15

 otherwise

1.14

 alternatively

1.14

chid

1.13

ifice

1.11

Else

1.10

acular

1.09

Activations Density 0.183%

phrases related to legal terms and violations

phrases that indicate conditional scenarios or alternatives

No Comments

No Known Activations