INDEX

Explanations

references to violence, executions, and their context in narrative scenarios

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

Verbs and phrases describing harmful or violent human actions (like killing, trading slaves, torture) or participation in morally reprehensible activities.

eleuther_acts_top20 · claude-3-7-sonnet-20250219 Triggered by @lu-christina

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_131k/average_l0_114

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 незавершена

-0.57

 Administrativna

-0.45

Viited

-0.45

Abraços

-0.45

demment

-0.45

ymce

-0.44

日閲覧

-0.44

omiast

-0.44

accessible

-0.43

Initializable

-0.43

POSITIVE LOGITS

 murder

0.43

 membunuh

0.39

 killing

0.39

 acts

0.38

 murders

0.38

murder

0.38

 remor

0.37

 crimes

0.37

Murder

0.37

 endforeach

0.36

Activations Density 0.057%

references to violence, executions, and their context in narrative scenarios

Verbs and phrases describing harmful or violent human actions (like killing, trading slaves, torture) or participation in morally reprehensible activities.

No Comments

No Known Activations