© Neuronpedia 2026

Privacy & Terms Blog GitHub Slack Twitter Contact

Neuronpedia

Natural Language

NEW Assistant AxisNEW Circuit TracerUPDATESteer SAE Evals ExportsAPI Community Blog Privacy & Terms Contact

LLAMA3.1-8B-IT · 15-RESID-POST-AA · 100878 ｜ Neuronpedia

Home
Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
Llama3.1-8B-IT
Resid Post - 131k
15-RESID-POST-AA
100878

INDEX

Explanations

spell

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

 Oxygen

-0.08

txt

-0.08

 Baker

-0.07

 Diane

-0.07

 machen

-0.07

 Jacobs

-0.07

ax

-0.07

 audiences

-0.07

 cuatro

-0.07

.business

-0.07

POSITIVE LOGITS

 spell

0.11

 Spell

0.10

Spell

0.09

 spelled

0.08

 implicitly

0.07

 smell

0.07

_spell

0.07

召

0.07

لل

0.06

ELL

0.06

Activations Density 0.008%

No Known Activations