INDEX

Explanations

numbers and sequences such as "4-3-2-1" or "3-3-7"

oai_token-act-pair · gpt-3.5-turbo

numeric values associated with shapes and configurations

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

illion

-0.75

 contaminants

-0.74

 Lifetime

-0.72

âĿ

-0.71

llah

-0.70

 tears

-0.70

verified

-0.70

reviewed

-0.69

 verified

-0.67

netflix

-0.67

POSITIVE LOGITS

 formations

1.36

 tactics

1.25

 playbook

1.23

 offense

1.20

 scheme

1.20

 philosophies

1.18

 schemes

1.17

 Offense

1.13

 formation

1.11

 philosophy

1.11

Activations Density 0.386%

numbers and sequences such as "4-3-2-1" or "3-3-7"

numeric values associated with shapes and configurations

No Comments

No Known Activations