INDEX

Explanations

names related to legal or authority positions

oai_token-act-pair · gpt-3.5-turbo

references to legal issues and criminal activities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 flavours

-1.06

 colours

-0.89

 flavour

-0.89

 favourites

-0.88

 favour

-0.86

 humour

-0.84

 behaviours

-0.83

 organisers

-0.83

 apologise

-0.82

 recognise

-0.82

POSITIVE LOGITS

FBI

1.01

 Congressional

0.99

ATF

0.96

 Reagan

0.94

FBI

0.94

DEA

0.93

 Federal

0.91

IRS

0.91

 Arkansas

0.88

Federal

0.88

Activations Density 1.019%

names related to legal or authority positions

references to legal issues and criminal activities

No Comments

No Known Activations