INDEX

Explanations

actions involving physical violence or harm

oai_token-act-pair · gpt-3.5-turbo

mentions of individuals involved in violent or criminal incidents

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ISTORY

-0.75

 renaissance

-0.71

 consultants

-0.70

iencies

-0.63

 reinvest

-0.62

lance

-0.61

 memoir

-0.61

reviewed

-0.59

PDATED

-0.59

OUP

-0.59

POSITIVE LOGITS

 unconscious

1.08

 sleeping

0.85

 naked

0.85

 senseless

0.83

 nearby

0.81

 unprotected

0.81

 injuring

0.81

 unaware

0.80

 helpless

0.79

 asleep

0.79

Activations Density 0.543%

actions involving physical violence or harm

mentions of individuals involved in violent or criminal incidents

No Comments

No Known Activations