INDEX

Explanations

mentions of team members or people working together

oai_token-act-pair · gpt-3.5-turbo

references to teammates in sports or group settings

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

atra

-0.81

brow

-0.78

osc

-0.76

skinned

-0.71

warning

-0.69

isen

-0.67

tarians

-0.66

olia

-0.66

ifax

-0.65

ingo

-0.65

POSITIVE LOGITS

 teammate

1.17

 teammates

1.09

mates

1.00

ubs

0.91

 mates

0.83

peak

0.78

 colleagues

0.78

 nicknamed

0.75

AVG

0.70

":[{"

0.69

Activations Density 0.015%

mentions of team members or people working together

references to teammates in sports or group settings

No Comments

No Known Activations