INDEX

Explanations

informational prompts asking for help or tips regarding law enforcement cases

oai_token-act-pair · gpt-3.5-turbo

phrases related to requests for information about crimes or suspects

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

 intest

-0.71

irements

-0.68

ãģł

-0.65

 textbooks

-0.64

 proofs

-0.63

 fulfil

-0.63

 autop

-0.63

ateral

-0.63

addle

-0.61

 patents

-0.61

POSITIVE LOGITS

 Crime

1.12

 Crim

1.12

Anyone

1.01

FBI

0.97

 Detective

0.94

Police

0.91

 Investigator

0.90

 homicide

0.89

 detectives

0.89

Investigators

0.86

Activations Density 0.088%

informational prompts asking for help or tips regarding law enforcement cases

phrases related to requests for information about crimes or suspects

No Comments

No Known Activations