INDEX

Explanations

mentions of criminal suspects or persons of interest in law enforcement contexts

oai_token-act-pair · gpt-3.5-turbo

repeated references to a suspect in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

issance

-0.76

 Remastered

-0.74

galitarian

-0.73

atism

-0.73

 stead

-0.66

Morning

-0.65

 refreshing

-0.64

ummer

-0.63

UNCH

-0.62

unch

-0.62

POSITIVE LOGITS

 suspects

0.82

 apprehended

0.77

 arrested

0.77

 suspect

0.77

 kidn

0.76

 sentenced

0.76

FBI

0.75

 gunman

0.74

eering

0.74

 perpetrator

0.73

Activations Density 0.035%

mentions of criminal suspects or persons of interest in law enforcement contexts

repeated references to a suspect in various contexts

No Comments

No Known Activations