INDEX

Explanations

words related to legal terms and offenses

oai_token-act-pair · gpt-3.5-turbo

terms related to legal or formal accusations and offenses

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 brig

-0.73

rh

-0.72

bal

-0.66

è

-0.66

po

-0.64

boo

-0.62

 goblin

-0.62

pig

-0.61

bur

-0.61

bas

-0.61

POSITIVE LOGITS

ession

4.47

essions

3.59

ESSION

2.20

essional

1.89

essing

1.57

essor

1.40

essors

1.37

essed

1.37

ettings

1.23

ECTION

1.16

Activations Density 0.012%

words related to legal terms and offenses

terms related to legal or formal accusations and offenses

No Comments

No Known Activations