© Neuronpedia 2026

Privacy & Terms Blog GitHub Slack Twitter Contact

Neuronpedia

APIAssistant AxisNEW Circuit TracerNEW Steer SAE Evals Exports Community Blog Privacy & Terms Contact

LLAMA3.1-8B-IT · 15-RESID-POST-AA · 115115 ｜ Neuronpedia

Home
Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
Llama3.1-8B-IT (Instruct)
Resid Post - 131k
15-RESID-POST-AA
115115

INDEX

Explanations

punctuation

np_max-act · gemini-2.0-flash

New Auto-Interp

Top Features by Cosine Similarity

Configuration

andyrdt/saes-llama-3.1-8b-instruct/resid_post_layer_15/trainer_1

Dataset (Dashboard)

Various

Features

131,072

Data Type

float32

Hook Name

blocks.15.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

Show PlotsShow ExplanationShow ActivationsShow Test FieldShow SteerShow Link

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 serves

-0.07

 shack

-0.06

 Advertising

-0.06

 excuses

-0.06

flare

-0.06

 insanların

-0.06

CALL

-0.06

 Ruth

-0.06

 cố

-0.06

slash

-0.06

POSITIVE LOGITS

/il

0.07

 doPost

0.06

ゝ

0.06

%"

0.06

usted

0.06

＜

0.06

izable

0.06

нова

0.06

ίναι

0.06

0.06

Activations Density 0.107%

No Known Activations