INDEX

Explanations

changes

np_max-act · gemini-2.0-flash

The neuron activates on instances of “commands” (and similar control-oriented verbs) marking where one party issues orders or obligations.

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

The neuron fires on controlling, imperative language—specifically words that issue commands or obligations (e.g. “commands,” “must,” etc.).

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

(targetEntity

-0.07

 konk

-0.07

anik

-0.07

 эффек

-0.07

 cudd

-0.07

=(

-0.06

 тепер

-0.06

inqu

-0.06

_PARAM

-0.06

rak

-0.06

POSITIVE LOGITS

 battleground

0.06

wig

0.06

,label

0.06

owej

0.06

	Point

0.06

Una

0.06

.BufferedReader

0.06

 gateway

0.06

 restraining

0.06

 Barack

0.06

Activations Density 0.000%

changes

The neuron activates on instances of “commands” (and similar control-oriented verbs) marking where one party issues orders or obligations.

The neuron fires on controlling, imperative language—specifically words that issue commands or obligations (e.g. “commands,” “must,” etc.).

No Comments

No Known Activations