INDEX

Explanations

phrases related to government surveillance or intelligence programs

oai_token-act-pair · gpt-3.5-turbo

references to surveillance practices and government monitoring

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Revival

-0.89

 Parables

-0.87

 Horses

-0.85

åħī

-0.82

 Wheels

-0.82

Coach

-0.81

pring

-0.80

 Rally

-0.77

 Riding

-0.77

 Miko

-0.75

POSITIVE LOGITS

 spying

1.57

 surveillance

1.55

illance

1.44

 espionage

1.40

NSA

1.36

spy

1.36

NSA

1.35

 malware

1.33

 cybersecurity

1.30

 hacking

1.30

Activations Density 1.251%

phrases related to government surveillance or intelligence programs

references to surveillance practices and government monitoring

No Comments

No Known Activations