INDEX

Explanations

mentions of individuals who held specific previous professional titles or roles

oai_token-act-pair · gpt-3.5-turbo

references to individuals' past roles or occupations

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

otle

-0.83

ammy

-0.82

utterstock

-0.78

andise

-0.74

antics

-0.74

allery

-0.72

matic

-0.71

acht

-0.70

achus

-0.70

aris

-0.69

POSITIVE LOGITS

 Yugoslav

1.00

 Yugoslavia

0.87

 Soviet

0.80

 employee

0.78

 pupil

0.76

 classmate

0.75

foundland

0.74

 colleague

0.71

 member

0.70

 employer

0.70

Activations Density 0.027%

mentions of individuals who held specific previous professional titles or roles

references to individuals' past roles or occupations

No Comments

No Known Activations