INDEX

Explanations

specific instances of people serving in various roles or positions

oai_token-act-pair · gpt-3.5-turbo

instances of the word "served."

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

issues

-0.83

orph

-0.82

pled

-0.69

eros

-0.68

destruct

-0.68

opic

-0.66

 fragile

-0.62

ylon

-0.62

wise

-0.61

illion

-0.61

POSITIVE LOGITS

 served

1.07

 Serving

1.00

 Restaur

0.94

 SERV

0.89

 serving

0.88

Serv

0.87

 kitchens

0.84

 Serve

0.83

 Entered

0.81

 meals

0.80

Activations Density 0.015%

specific instances of people serving in various roles or positions

instances of the word "served."

No Comments

No Known Activations