INDEX

Explanations

word tokens related to legal or legislative matters or processes

oai_token-act-pair · gpt-3.5-turbo

references to constitutional amendments

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

 Nemesis

-0.70

 Fren

-0.67

ocked

-0.61

 Stud

-0.60

Stud

-0.59

veyard

-0.59

whel

-0.58

etr

-0.58

pd

-0.58

este

-0.57

POSITIVE LOGITS

 amendment

3.79

 amendments

2.80

 Amendment

2.69

 Amendments

2.22

endment

1.95

 amend

1.76

 amended

1.63

 Amend

1.58

 alteration

1.23

 exemption

1.20

Activations Density 0.016%

word tokens related to legal or legislative matters or processes

references to constitutional amendments

No Comments

No Known Activations