INDEX

Explanations

the word "bow" with variations such as "bowed" and "bows"

oai_token-act-pair · gpt-3.5-turbo

instances of the word "bow" 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

Negative Logits

oci

-0.79

inst

-0.75

area

-0.71

papers

-0.71

sym

-0.71

 Helsinki

-0.71

MEM

-0.70

irtual

-0.69

 INST

-0.67

housing

-0.67

POSITIVE LOGITS

bow

4.07

 bows

3.01

Bow

2.01

Bow

1.99

 bowed

1.94

bow

1.77

 Bows

1.66

bows

1.38

 bowling

1.30

 arrow

1.29

Activations Density 0.017%

the word "bow" with variations such as "bowed" and "bows"

instances of the word "bow" in various contexts

No Known Activations