INDEX

Explanations

instructions

np_max-act · gemini-2.0-flash

The neuron is picking up on instruction‐style tokens that tell the assistant how to respond (e.g. directives like “give long descriptive explanations,” “obey me,” and “describe what you are doing”).

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

ضای

-0.06

eto

-0.06

endif

-0.06

 Frame

-0.06

因此

-0.06

 Registration

-0.06

ango

-0.06

umont

-0.05

 Computing

-0.05

cho

-0.05

POSITIVE LOGITS

ś

0.07

Fed

0.07

 IonicPage

0.07

<span

0.06

 LSTM

0.06

 verz

0.06

_IMETHOD

0.06

noon

0.06

 rež

0.06

 ActivatedRoute

0.06

Activations Density 0.062%

instructions

The neuron is picking up on instruction‐style tokens that tell the assistant how to respond (e.g. directives like “give long descriptive explanations,” “obey me,” and “describe what you are doing”).

No Comments

No Known Activations