INDEX

Explanations

phrases or terms related to "Top" rankings or lists

oai_token-act-pair · gpt-3.5-turbo

headings or titles highlighted in a list format

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 amen

-0.79

 unle

-0.68

 compassion

-0.68

 corridor

-0.66

 rebel

-0.66

 severity

-0.64

 servant

-0.64

 grievance

-0.63

 suffering

-0.63

 decree

-0.62

POSITIVE LOGITS

Top

3.76

TOP

2.37

top

2.26

Top

2.25

Bottom

1.92

TOP

1.81

top

1.69

 Bottom

1.46

tops

1.45

bottom

1.43

Activations Density 0.010%

phrases or terms related to "Top" rankings or lists

headings or titles highlighted in a list format

No Comments

No Known Activations