© Neuronpedia 2026

Privacy & Terms Blog GitHub Slack Twitter Contact

Neuronpedia

Jacobian LensNEW

Natural Language

NEW Assistant AxisNEW Circuit TracerUPDATESteer SAE Evals Exports interp-engineNEWAPI Community Blog Privacy & Terms Contact

QWEN3-32B · 32-RESID-BATCHTOPK-65K · 5168 ｜ Neuronpedia

Home
Qwen3-32B
32-RESID-BATCHTOPK-65K
5168

INDEX

Explanations

shape or board grid

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

New Auto-Interp

Top Features by Cosine Similarity

Configuration

adamkarvonen/qwen3-32b-saes/saes_Qwen_Qwen3-32B_batch_top_k/resid_post_layer_32

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

No Configuration Found

Embeds

Show PlotsShow ExplanationShow ActivationsShow Test FieldShow SteerShow Link

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 thuyáº¿t

-0.10

åĪĳ

-0.10

é¥¯

-0.09

è´¯

-0.09

æµģ

-0.09

æ®·

-0.08

åıįè¿ĩæĿ¥

-0.08

egg

-0.08

Ð´Ð°ÑĢ

-0.08

æĩĪ

-0.08

POSITIVE LOGITS

 square

0.12

square

0.11

(square

0.11

.square

0.11

-square

0.10

 squares

0.10

æĸ¹å½¢

0.10

è¿ĻåĿĹ

0.10

ç«ĭæĸ¹

0.09

 cube

0.09

Activations Density 0.095%

No Known Activations