INDEX

Explanations

references to unethical behavior or wrongdoing, particularly in official roles

oai_token-act-pair · gpt-3.5-turbo

references to various forms of misconduct

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

etically

-0.90

NetMessage

-0.79

osphere

-0.76

etics

-0.75

pop

-0.73

izen

-0.73

istan

-0.72

iku

-0.69

etic

-0.67

brush

-0.67

POSITIVE LOGITS

 misconduct

1.10

 wrongdoing

1.00

onduct

1.00

 unfocusedRange

0.92

 allegations

0.88

 disclosures

0.86

 malf

0.85

 disclosure

0.76

ionage

0.75

 dealings

0.74

Activations Density 0.043%

references to unethical behavior or wrongdoing, particularly in official roles

references to various forms of misconduct

No Comments

No Known Activations