INDEX

Explanations

Called by name

np_max-act · gemini-2.0-flash

The neuron detects template‐style instructions that specify how the character should address the user—e.g. lines telling the model “calls user by any name introduced by user.”

oai_token-act-pair · o4-mini Triggered by @xinyanhu8

New Auto-Interp

Configuration

andyrdt/saes-llama-3.1-8b-instruct/resid_post_layer_11/trainer_1

Dataset (Dashboard)

Various

Features

131,072

Data Type

float32

Hook Name

blocks.11.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

uden

-0.07

 لن

-0.06

 reservations

-0.06

 nursing

-0.06

	BufferedReader

-0.06

na

-0.06

 zwei

-0.06

 outputs

-0.06

madan

-0.06

 purchasing

-0.06

POSITIVE LOGITS

�

0.07

_UTF

0.07

_RPC

0.06

April

0.06

surname

0.06

oti

0.06

 beau

0.06

<Scalar

0.06

 alors

0.06

ald

0.06

Activations Density 0.003%

Called by name

The neuron detects template‐style instructions that specify how the character should address the user—e.g. lines telling the model “calls user by any name introduced by user.”

No Comments

No Known Activations