INDEX

Explanations

references to eating behaviors or actions related to food consumption

oai_token-act-pair · gpt-3.5-turbo

instances of the word "eating" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Mehran

-0.76

vals

-0.76

AU

-0.76

 Spectre

-0.76

 Boll

-0.68

lv

-0.65

 Static

-0.64

bors

-0.63

meier

-0.62

Revolution

-0.62

POSITIVE LOGITS

 eating

3.48

 Eating

2.94

eating

2.34

eat

2.29

ate

2.07

 eats

1.99

 eaten

1.93

 consuming

1.91

 swallowing

1.88

Eat

1.81

Activations Density 0.009%

references to eating behaviors or actions related to food consumption

instances of the word "eating" in various contexts

No Comments

No Known Activations