INDEX

Explanations

to

np_max-act · gemini-2.0-flash

The neuron detects the infinitive “to” in instructional questions (e.g. the “to” in “how to …”).

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

This neuron detects user requests for instructions — i.e., "how to" / how-do-I style how-to questions.

oai_token-act-pair · gpt-5-mini Triggered by @vetterc0

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 slic

-0.07

 ware

-0.06

Bet

-0.06

whel

-0.06

_post

-0.06

 memoria

-0.06

bum

-0.06

trap

-0.06

.strict

-0.06

 disturbed

-0.06

POSITIVE LOGITS

上了

0.07

 пит

0.07

 Στα

0.06

OBJ

0.06

 kesin

0.06

�

0.06

外

0.06

ق

0.06

 succes

0.06

 Làm

0.06

Activations Density 0.033%

to

The neuron detects the infinitive “to” in instructional questions (e.g. the “to” in “how to …”).

This neuron detects user requests for instructions — i.e., "how to" / how-do-I style how-to questions.

No Comments

No Known Activations