INDEX

Explanations

instances where someone is being accused of deception

oai_token-act-pair · gpt-3.5-turbo

instances of the word "lying" to indicate dishonesty or falsehood

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Ultra

-0.77

ugal

-0.72

FN

-0.72

ORE

-0.72

entry

-0.71

era

-0.71

Effective

-0.71

aud

-0.71

ISO

-0.70

aldi

-0.70

POSITIVE LOGITS

 lying

0.94

 horizont

0.89

 liar

0.80

lie

0.79

 skelet

0.78

 seiz

0.77

 lied

0.77

 pills

0.75

 mortg

0.74

 camoufl

0.73

Activations Density 0.006%

instances where someone is being accused of deception

instances of the word "lying" to indicate dishonesty or falsehood

No Comments

No Known Activations