INDEX

Explanations

mentions of various forms and contexts of harassment

oai_token-act-pair · gpt-3.5-turbo

references to harassment, particularly sexual harassment

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

iol

-0.68

 Bran

-0.67

ym

-0.67

ynt

-0.66

alog

-0.65

omer

-0.65

lean

-0.65

 Ital

-0.65

 OPEC

-0.65

Ãĥ

-0.64

POSITIVE LOGITS

 harassment

3.51

 harass

2.40

 harassing

2.17

 harassed

2.01

assment

1.96

 intimidation

1.78

 bullying

1.66

 stalking

1.62

 discrimination

1.58

 abuse

1.48

Activations Density 0.018%

mentions of various forms and contexts of harassment

references to harassment, particularly sexual harassment

No Comments

No Known Activations