INDEX

Explanations

phrases related to recommendations or preferences

oai_token-act-pair · gpt-3.5-turbo

phrases related to recommendations or lists of favorites

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 observers

-0.85

 clinicians

-0.83

ensing

-0.82

 analysts

-0.82

 investors

-0.81

 shareholders

-0.80

 enrollment

-0.80

 investigators

-0.79

iminary

-0.76

 supplementation

-0.76

POSITIVE LOGITS

 Strawberry

1.01

 Banana

0.90

 Kinnikuman

0.89

 Cinnamon

0.89

 Carib

0.87

 Vanilla

0.87

 Cowboy

0.85

 Tropical

0.84

 Cinderella

0.83

 Grizz

0.83

Activations Density 0.387%

phrases related to recommendations or preferences

phrases related to recommendations or lists of favorites

No Comments

No Known Activations