INDEX

Explanations

advice and concern

np_max-act · gemini-2.0-flash

This neuron detects polite, advisory phrases offering guidance (e.g. “It’s important to…,” “you can…,” “approach…,” “express your concerns,” 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

_Params

-0.07

ठन

-0.06

invisible

-0.06

 Sınıf

-0.06

jumlah

-0.06

 herpes

-0.06

rec

-0.06

 propulsion

-0.06

	info

-0.06

rá

-0.06

POSITIVE LOGITS

此

0.07

=w

0.06

{l

0.06

others

0.06

hound

0.06

стит

0.06

 getSession

0.06

 researching

0.06

 poor

0.06

 Thousands

0.06

Activations Density 0.051%

advice and concern

This neuron detects polite, advisory phrases offering guidance (e.g. “It’s important to…,” “you can…,” “approach…,” “express your concerns,” etc.).

No Comments

No Known Activations