INDEX

Explanations

strongly negative and offensive language

oai_token-act-pair · gpt-3.5-turbo

negative descriptors related to immorality and offensive behavior

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

ership

-0.82

arat

-0.74

essor

-0.73

ILA

-0.73

inoa

-0.72

OTO

-0.71

arist

-0.69

 Airl

-0.68

oglu

-0.68

eer

-0.68

POSITIVE LOGITS

 vile

0.82

 despicable

0.81

 disgusting

0.81

 slurs

0.79

 hateful

0.77

 racist

0.76

 racists

0.75

 misogyn

0.75

soever

0.73

 creatures

0.70

Activations Density 0.029%

strongly negative and offensive language

negative descriptors related to immorality and offensive behavior

No Comments

No Known Activations