INDEX

Explanations

official job titles or descriptions

oai_token-act-pair · gpt-3.5-turbo

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

NetMessage

-0.83

icides

-0.79

 puppies

-0.71

 dinosaurs

-0.70

 kittens

-0.70

 monsters

-0.69

 murderers

-0.69

 calves

-0.69

rooms

-0.69

 TAMADRA

-0.69

POSITIVE LOGITS

 source

1.49

 spokesman

1.27

 spokeswoman

1.23

 spokesperson

1.23

 insider

1.16

 aide

1.11

 representative

1.07

 diplomat

1.06

 senior

1.06

 staffer

1.03

Activations Density 0.191%

official job titles or descriptions

No Comments

No Known Activations