INDEX

Explanations

promoting

np_max-act · gemini-2.0-flash

The neuron fires on key words and phrases that name or introduce unsafe‐content categories (e.g. “sexual…arouse,” “promotes,” “depicts,” “incites,” “self-harm,” “violence,” etc.), effectively marking tokens that specify policy violation types.

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

άνα

-0.07

ackages

-0.07

ITS

-0.06

	items

-0.06

.buffer

-0.06

ERT

-0.06

Server

-0.06

 dimensions

-0.06

monitor

-0.06

isateur

-0.06

POSITIVE LOGITS

，一

0.08

ebi

0.07

rtype

0.07

 starttime

0.07

dığını

0.06

Tub

0.06

 eBook

0.06

adolu

0.06

 Hoover

0.06

 quarterbacks

0.06

Activations Density 0.016%

promoting

The neuron fires on key words and phrases that name or introduce unsafe‐content categories (e.g. “sexual…arou­se,” “promotes,” “depicts,” “incites,” “self-harm,” “violence,” etc.), effectively marking tokens that specify policy violation types.

No Comments

No Known Activations

The neuron fires on key words and phrases that name or introduce unsafe‐content categories (e.g. “sexual…arouse,” “promotes,” “depicts,” “incites,” “self-harm,” “violence,” etc.), effectively marking tokens that specify policy violation types.