INDEX

Explanations

terms related to bullying

oai_token-act-pair · gpt-3.5-turbo

references to bullying and its related terms

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

Num

-0.76

Num

-0.74

lime

-0.72

ources

-0.70

apache

-0.70

 Lumin

-0.69

 Recon

-0.69

Hal

-0.68

 Interior

-0.67

 Illum

-0.67

POSITIVE LOGITS

 bullying

3.31

 bullies

3.12

 bully

3.04

 bullied

2.94

bull

1.52

 blackmail

1.40

 classmate

1.35

Bull

1.25

 homophobic

1.24

 trolls

1.24

Activations Density 0.023%

terms related to bullying

references to bullying and its related terms

No Comments

No Known Activations