INDEX

Explanations

statements or references related to causing offense

oai_token-act-pair · gpt-3.5-turbo

language related to causing offense or being offended

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Monitor

-0.77

 Phase

-0.70

omed

-0.69

 Helm

-0.68

 Canaver

-0.68

packed

-0.66

 Maid

-0.65

Mech

-0.65

Phase

-0.65

 Motors

-0.65

POSITIVE LOGITS

 offending

3.38

 offended

3.03

 offend

2.94

 insulted

1.87

 angered

1.49

 infring

1.46

 offence

1.45

 insult

1.39

 blasp

1.36

 insults

1.31

Activations Density 0.026%

statements or references related to causing offense

language related to causing offense or being offended

No Comments

No Known Activations