INDEX

Explanations

phrases related to investigations or allegations

oai_token-act-pair · gpt-3.5-turbo

repeated phrases that indicate allegations of wrongdoing

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

uristic

-0.90

ieties

-0.76

ertodd

-0.74

hap

-0.71

 folds

-0.70

heses

-0.68

nets

-0.68

Tokens

-0.68

 partName

-0.68

keys

-0.68

POSITIVE LOGITS

 inacc

0.95

 wrongdoing

0.88

 harassment

0.80

 misconduct

0.75

 misinformation

0.74

 discrimination

0.71

 criminality

0.71

 violence

0.71

 vandalism

0.70

 foul

0.70

Activations Density 0.158%

phrases related to investigations or allegations

repeated phrases that indicate allegations of wrongdoing

No Comments

No Known Activations