INDEX

Explanations

phrases related to actions being blocked or halted

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate actions or conditions imposed by external authorities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

raq

-0.87

 largeDownload

-0.86

apest

-0.83

empl

-0.82

aughtered

-0.77

verty

-0.77

yssey

-0.76

alysed

-0.75

redits

-0.74

enth

-0.72

POSITIVE LOGITS

 virtue

1.12

 sheer

0.90

 successive

0.84

laws

0.83

 inaction

0.80

 fiat

0.80

 default

0.78

 bureaucracy

0.76

 retreating

0.74

 authorities

0.73

Activations Density 0.174%

phrases related to actions being blocked or halted

phrases that indicate actions or conditions imposed by external authorities

No Comments

No Known Activations