INDEX

Explanations

instances of the word "cans" combined with different contexts and modifiers

oai_token-act-pair · gpt-3.5-turbo

references to cans, particularly food containers

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

Tut

-0.80

iae

-0.78

yna

-0.69

rick

-0.67

ring

-0.67

 Guest

-0.66

 knight

-0.66

edge

-0.64

agic

-0.64

 Ward

-0.63

POSITIVE LOGITS

 cans

3.91

 bottles

1.77

 canned

1.67

 crates

1.42

 bins

1.37

 cups

1.34

 cones

1.33

 jars

1.32

 containers

1.32

 packets

1.24

Activations Density 0.019%

instances of the word "cans" combined with different contexts and modifiers

references to cans, particularly food containers

No Comments

No Known Activations