INDEX

Explanations

words related to inappropriate or offensive behavior

oai_token-act-pair · gpt-3.5-turbo

terms related to indecency and inappropriate behavior

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

grass

-0.77

 Bauer

-0.72

CI

-0.72

SourceFile

-0.67

oreal

-0.65

 Labs

-0.65

Sap

-0.64

 labs

-0.64

VM

-0.63

verified

-0.63

POSITIVE LOGITS

 indec

3.51

 indecent

2.80

 lewd

1.96

 disorderly

1.53

 obscene

1.53

 grop

1.37

 inco

1.24

 indisc

1.23

 decency

1.17

 obsc

1.12

Activations Density 0.030%

words related to inappropriate or offensive behavior

terms related to indecency and inappropriate behavior

No Comments

No Known Activations