INDEX

Explanations

The neuron is looking for mentions of alleged incidents or crimes

oai_token-act-pair · gpt-3.5-turbo

repeated mentions of the word "alleged."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

uden

-0.88

uda

-0.81

erie

-0.76

ð

-0.74

cair

-0.73

ada

-0.71

è£ħ

-0.71

_.

-0.70

utics

-0.70

haar

-0.69

POSITIVE LOGITS

 perpetrator

1.02

 wrongdoing

1.00

 violations

0.97

 perpetrators

0.95

 abuser

0.95

 rapist

0.95

 culprit

0.95

 collusion

0.91

 abuses

0.91

 assailant

0.91

Activations Density 0.079%

The neuron is looking for mentions of alleged incidents or crimes

repeated mentions of the word "alleged."

No Comments

No Known Activations