INDEX

Explanations

mentions of lying or related terms

oai_token-act-pair · gpt-3.5-turbo

expressions related to dishonesty and deceit

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

aldi

-0.85

ugal

-0.76

joining

-0.76

orsi

-0.75

allows

-0.74

FN

-0.72

ains

-0.69

illed

-0.69

obs

-0.69

ategory

-0.69

POSITIVE LOGITS

 detector

1.01

uten

0.94

 detectors

0.77

 deceit

0.76

utenant

0.76

 vulner

0.74

 liar

0.74

 misrepresent

0.73

acies

0.73

 deceive

0.72

Activations Density 0.021%

mentions of lying or related terms

expressions related to dishonesty and deceit

No Comments

No Known Activations