穆再排尔·穆合塔尔

Portfolio & field notes — 2026

MACHINES THAT SEE.

机器如何看。

I build vision systems for drones and phones — and write about the distance between what a model sees and what we meant.

Nanjing, CN Portfolio 2023—2026 Class of 2028 · open to research
FIG. 01f / 1.4
Fig. 01 — Aperture, study 012026
Optical study / 2026 32.0603° N / 118.7969° E
Scroll to read

01 Selected work

Selected work

Four systems, one question: what does the model actually see?

All work

On error

A model does not fail loudly. It fails in the space between the label and the thing itself.
From “The 18.4% Problem” ↗
First page of the NeurIPS extended abstract PDF / 6 PP.

Research / NeurIPS 2026

Selective Cross-Modal Alignment

Why position grounding succeeds but physical reasoning fails.

6,000 controlled physics clips. Position is explicit; velocity is a manifold the language head ignores.

position / linear probe
R² 0.97
language grounding
100%
with prompting
+62%
A black-and-white photograph of Muzapar Muhtar setting up a camera and laptop on a tripod outdoors.
PL. 05 Plate — self, at work · Nanjing · fieldwork

03 About

A practice of looking closely.

I am a mechanical engineering undergraduate at NUAA, working on computer vision for autonomous systems. My research looks at selective cross-modal alignment in vision-language models. My engineering work lives on the edge: drones, phones, TensorRT.

Selected systems
04
Workshop paper
01
Class of
2028