INDEX

Explanations

instances of words related to legal or official terms, specifically related to seeking exceptions or special permissions

oai_token-act-pair · gpt-3.5-turbo

references to legal or religious exemptions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ebus

-0.75

 Xiang

-0.75

riz

-0.74

 Roses

-0.72

liam

-0.67

 Fury

-0.66

kr

-0.64

erker

-0.63

 Bras

-0.60

 Cardinal

-0.60

POSITIVE LOGITS

 exemptions

1.23

 exemption

1.08

 exempt

1.07

Reviewer

1.01

 exempted

0.94

untarily

0.85

exempt

0.84

 immunity

0.79

 minded

0.78

 amnesty

0.77

Activations Density 0.007%

instances of words related to legal or official terms, specifically related to seeking exceptions or special permissions

references to legal or religious exemptions

No Comments

No Known Activations