INDEX

Explanations

winning-related words and phrases

oai_token-act-pair · gpt-3.5-turbo

mentions of winning or success

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

 underground

-0.78

 arom

-0.76

 involuntary

-0.72

 serpent

-0.69

 acid

-0.68

 infusion

-0.67

 feral

-0.65

 contempl

-0.64

 antim

-0.63

 masked

-0.63

POSITIVE LOGITS

Win

4.00

win

2.15

WIN

2.13

Win

1.67

WIN

1.57

 Winning

1.50

winning

1.46

Winner

1.45

winner

1.35

win

1.26

Activations Density 0.013%

winning-related words and phrases

mentions of winning or success

No Comments

No Known Activations