INDEX

Explanations

mentions of people's former professional positions and affiliations

oai_token-act-pair · gpt-3.5-turbo

references to individuals' previous roles or positions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

anium

-0.83

llers

-0.81

Rs

-0.80

places

-0.78

abouts

-0.78

ombies

-0.77

tics

-0.77

houses

-0.76

fixes

-0.76

iths

-0.75

POSITIVE LOGITS

 staffer

1.17

 employee

1.15

 classmate

1.14

 aide

1.13

 member

1.11

 adviser

1.10

 prosecutor

1.09

 colleague

1.06

 deputy

1.02

 diplomat

1.02

Activations Density 0.076%

mentions of people's former professional positions and affiliations

references to individuals' previous roles or positions

No Comments

No Known Activations