INDEX

Explanations

mentions of actions or decisions that need to be taken based on a situation or circumstance

oai_token-act-pair · gpt-3.5-turbo

phrases indicating acknowledgement of wrongdoing or the need for action

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

IFrame

Link

Negative Logits

ipel

-0.69

aten

-0.64

 Dwell

-0.63

 millenn

-0.63

illi

-0.62

etta

-0.62

emetery

-0.62

bris

-0.61

 sshd

-0.61

 corridors

-0.60

POSITIVE LOGITS

 wrong

0.98

 differently

0.94

 rash

0.88

 else

0.88

wrong

0.88

 proactive

0.83

 naughty

0.81

else

0.78

 offensively

0.77

 stupid

0.76

Activations Density 0.065%

mentions of actions or decisions that need to be taken based on a situation or circumstance

phrases indicating acknowledgement of wrongdoing or the need for action

No Known Activations