INDEX

Explanations

phrases related to different forms and contexts of abuse, including financial abuse and historical abuse allegations

oai_token-act-pair · gpt-3.5-turbo

mentions of abuse in various contexts

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

Not in Any Lists

No Comments

Negative Logits

cil

-0.80

travel

-0.74

pard

-0.73

views

-0.71

izen

-0.70

 compr

-0.68

cise

-0.67

vision

-0.66

ript

-0.66

pai

-0.65

POSITIVE LOGITS

abuse

0.93

 inflicted

0.89

 perpetrated

0.87

 abusing

0.86

 abuse

0.83

 abused

0.77

 abusers

0.76

 victims

0.76

 Abuse

0.74

 abuses

0.73

Activations Density 0.037%

phrases related to different forms and contexts of abuse, including financial abuse and historical abuse allegations

mentions of abuse in various contexts

No Comments

No Known Activations