INDEX

Explanations

"Thank you", "Q:", "Ludwig Jungermann", "tag:"

np_acts-logits-general · gemini-2.5-flash-lite

This neuron activates on escaped Unicode or character‐reference sequences (e.g. “\u1234” or “'”), i.e. text representing characters via escape codes.

oai_token-act-pair · o4-mini

Triggered by @jyhe0408

New Auto-Interp

Configuration

google/gemma-scope-27b-pt-res/layer_10/width_131k

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

ządu

-2.36

铼

-2.34

atschappij

-2.34

ll

-2.30

罌

-2.20

and

-2.11

}=

-2.09

镭

-2.08

al

-2.06

 旅游

-2.05

POSITIVE LOGITS

1.95

🕗

1.90

 сейчас

1.87

1.84

 When

1.84

来到了

1.83

 quando

1.83

1.80

至於

1.78

ハ

1.77

Activations Density 0.004%

"Thank you", "Q:", "Ludwig Jungermann", "tag:"

This neuron activates on escaped Unicode or character‐reference sequences (e.g. “\u1234” or “'”), i.e. text representing characters via escape codes.

No Known Activations

"Thank you", "Q:", "Ludwig Jungermann", "tag:"

This neuron activates on escaped Unicode or character‐reference sequences (e.g. “\u1234” or “'”), i.e. text representing characters via escape codes.

No Known Activations

"Thank you", "Q:", "Ludwig Jungermann", "tag:"

This neuron activates on escaped Unicode or character‐reference sequences (e.g. “\u1234” or “&#x27;”), i.e. text representing characters via escape codes.

No Known Activations

"Thank you", "Q:", "Ludwig Jungermann", "tag:"

This neuron activates on escaped Unicode or character‐reference sequences (e.g. “\u1234” or “&#x27;”), i.e. text representing characters via escape codes.

No Known Activations

This neuron activates on escaped Unicode or character‐reference sequences (e.g. “\u1234” or “'”), i.e. text representing characters via escape codes.

This neuron activates on escaped Unicode or character‐reference sequences (e.g. “\u1234” or “'”), i.e. text representing characters via escape codes.