INDEX

Explanations

accusations or allegations against someone or something

oai_token-act-pair · gpt-3.5-turbo

phrases containing accusations or claims of wrongdoing

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

isSpecialOrderable

-0.91

chance

-0.72

Oracle

-0.72

iland

-0.67

erm

-0.67

Chance

-0.65

FIG

-0.64

thank

-0.64

mint

-0.64

allows

-0.64

POSITIVE LOGITS

 improper

0.91

 extortion

0.87

 illegal

0.82

 manslaughter

0.81

 paed

0.80

 blackmail

0.80

 poisoning

0.80

 illegally

0.80

 disproportionate

0.79

 unethical

0.77

Activations Density 0.278%

accusations or allegations against someone or something

phrases containing accusations or claims of wrongdoing

No Known Activations