INDEX

Explanations

descriptions of violent acts, particularly murders

oai_token-act-pair · gpt-3.5-turbo

references to murder

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ffic

-0.78

aque

-0.74

rium

-0.71

issue

-0.71

patch

-0.69

Scot

-0.66

Champ

-0.65

alach

-0.64

BuyableInstoreAndOnline

-0.64

uber

-0.64

POSITIVE LOGITS

 murdered

0.96

 slain

0.83

 murdering

0.80

 assassinated

0.78

nesday

0.77

 innocent

0.76

 innoc

0.75

pload

0.75

 spree

0.75

soever

0.75

Activations Density 0.050%

descriptions of violent acts, particularly murders

references to murder

No Comments

No Known Activations