INDEX

Explanations

politically related terms or positions, specifically focusing on the political left

oai_token-act-pair · gpt-3.5-turbo

references to left-wing political ideology or affiliations

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

Not in Any Lists

No Comments

Negative Logits

 meeting

-0.63

sol

-0.63

 session

-0.61

 instant

-0.60

ser

-0.60

 fodder

-0.60

 resort

-0.60

 yearly

-0.59

 annually

-0.59

 service

-0.59

POSITIVE LOGITS

left

3.66

Left

2.56

right

2.44

 left

1.83

 Left

1.81

pictured

1.53

Right

1.45

bottom

1.36

center

1.32

 RIGHT

1.20

Activations Density 0.008%

politically related terms or positions, specifically focusing on the political left

references to left-wing political ideology or affiliations

No Comments

No Known Activations