INDEX

Explanations

language related to threat and violence

oai_token-act-pair · gpt-4o-mini Triggered by @bot

death and misfortune

np_acts-logits-general · gemini-2.5-flash-lite

This neuron detects language that expresses, endorses, or calls for violent harm, death, or assassination of individuals or groups.

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Wikiseite

-0.52

Rüyada

-0.43

Enllaces

-0.43

énario

-0.41

 autorytatywna

-0.41

routeProvider

-0.41

 gyhoeddwyd

-0.40

存于互联网档案馆

-0.40

 مواليد

-0.39

 ràng

-0.39

POSITIVE LOGITS

 Wikimedijinoj

0.66

UserScript

0.52

 oprot

0.51

windowFixed

0.49

uxxxx

0.48

GT

0.43

Gt

0.42

TypedDataSet

0.40

EndContext

0.40

eh

0.40

Activations Density 0.088%

language related to threat and violence

death and misfortune

This neuron detects language that expresses, endorses, or calls for violent harm, death, or assassination of individuals or groups.

No Comments

No Known Activations

language related to threat and violence

death and misfortune

This neuron detects language that expresses, endorses, or calls for violent harm, death, or assassination of individuals or groups.

No Comments

No Known Activations