INDEX

Explanations

terms related to terrorism or violent acts, specifically related to hijacking

oai_token-act-pair · gpt-3.5-turbo

references to hijackers and their actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Rite

-0.85

-0.76

ç¥ŀ

-0.72

 Angels

-0.70

 Teachers

-0.68

Downloadha

-0.67

remlin

-0.67

âĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢ

-0.65

 Habit

-0.65

ãģ®éŃĶ

-0.65

POSITIVE LOGITS

hij

1.22

 hijacked

0.92

ackers

0.90

inx

0.87

acking

0.85

intosh

0.84

acker

0.81

eering

0.79

ibl

0.78

acket

0.73

Activations Density 0.007%

terms related to terrorism or violent acts, specifically related to hijacking

references to hijackers and their actions

No Comments

No Known Activations