NTH

Demonstration of MANA: Multi-perspective Alignment of NPC Agents

AuthorsAmir Tahmasbi, Keshav Singh, Shaheer Raza, Owen Mei, Jason Wu

AffiliationsPurdue University, West Lafayette, Indiana, USA

November 19, 2026 2 min read
Watch on YouTube
The one-line take

MANA lets game developers step into an NPC’s shoes and demonstrate how it should behave, then lets the character act on its own.

Key results

20
Action logging interval

The system records embodied actions and nearby context every 20 ticks.

2
Demonstration duration

Roughly 2 minutes of in-game play shaped each character's behavior.

What the paper found

MANA is a Minecraft mod that lets developers author non-player character behavior by taking control of an NPC and demonstrating actions and dialogue in the game world. It records speech alongside embodied actions and nearby context, sampling actions every 20 ticks. A large language model distills those demonstrations into a character persona and an interpretable policy with spatial, sequencing, and persistence rules; the NPC then uses that profile to act and speak autonomously. In the demonstration, roughly 2 minutes of play shaped Fletcher into a confident hunter and Farmer into a cautious character who flees threats, builds a shelter, and returns to it at night. MANA also shows cross-modal generalization: action-only demonstrations can inform later dialogue, while learned personas remain apparent in NPC-to-NPC conversations. The contribution is an embodied, in-situ alternative to manually scripting or prompting NPC behavior, illustrated through contrasting characters and new situations rather than quantified benchmark results.

Original abstract

We present MANA (Multi-perspective Alignment of NPC Agents), an interactive approach for authoring NPC behavior through embodied demonstration. We show MANA in the context of Minecraft, where a developer aligns NPCs while playing the game itself. Activating MANA lets the developer dynamically switch into the point of view (POV) of any character and demonstrate, in-situ, how it would act given the surrounding context, through both movement and dialogue. MANA then distills each demonstration into a generalized policy that maps in-game context to actions. Finally, once the developer switches to another NPC or exits MANA mode, this generalized policy drives the NPC on its own. We demonstrate MANA by aligning two initially generic Minecraft NPCs: Fletcher and Farmer to act aggressively and cowardly, respectively. After roughly 2 minutes of in-game play, each character exhibits consistent and believable behaviors across new, unseen scenarios.

Read the original paper

More in Embodied AI

Browse all 69 papers →
01Embodied Ai

Kimodo-Policy: Learning Kinematic Action Models for Humanoid Loco-Manipulation

Yunheng Wang, Jiaxi Zhang, Wei Lu, Yuetong Fang, Taowen Wang, Yixiao Feng, Shuning Zhang, Junzhe Xu, Yichi Wang, Zizhao Yuan, Kailun Huang, Lusong Li, Mingqian Zhou, Zecui Zeng, Renjing Xu

Kimodo-Policy turns a pretrained whole-body motion generator into a control policy for humanoid robots that walk and manipulate objects.

Read analysis
03Embodied Ai

JustRUA: A Controlled Study of GPT-6 Robot-Use Agents

Zefan Wang, Yufan Dang, Xuanwei Shi, Mingqi Gao, Cheng Yin, SikYuen Tam, Hanyu Liu, Yasheng Wang, Yaxi Lu, Songyang Zhang, Xin Cong, Yuan Yao

A leaner harness helps a GPT-driven robot perform much better in simulation, showing that agent design choices can matter as much as learning.

Read analysis