INDEX

Explanations

and, comma

np_max-act · gemini-2.0-flash

coordinated phrases where “health” is linked to another term, especially the “and” or adjacent punctuation connecting them.

oai_token-act-pair · gpt-5 Triggered by @liuliujiujiu05

New Auto-Interp

Configuration

mwhanna/qwen3-4b-transcoders/layer_23.safetensors

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

163,840

Data Type

float32

Hook Name

blocks.23.mlp.hook_in

Architecture

transcoder

Context Size

8,192

Dataset

monology/pile-uncopyrighted

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

tractor

-0.26

oui

-0.25

pcodes

-0.23

éļĲå½¢

-0.23

ypad

-0.23

 blush

-0.23

çĵ£

-0.23

åĵªå®¶

-0.23

-pad

-0.23

ocratic

-0.22

POSITIVE LOGITS

 health

0.30

 safety

0.29

åģ¥åº·

0.29

çĶŁæ´»

0.29

èº«ä½ĵåģ¥åº·

0.28

çĶŁæ´»çļĦ

0.28

health

0.27

amina

0.26

 integrity

0.26

 Const

0.25

Activations Density 0.004%

and, comma

coordinated phrases where “health” is linked to another term, especially the “and” or adjacent punctuation connecting them.

No Comments

No Known Activations