CV

Research interests

Robotics and embodied perception, cross-view geo-localization (drone-to-satellite matching) and visual navigation in GPS-denied environments, robust computer vision under adverse weather and corruption, and efficient / on-device (Edge) AI: model compression and distillation for real-time inference on aerial platforms.

Education

B.Sc. in Artificial Intelligence, VNUHCM, University of Science (HCMUS), Sep 2023 - Present
K23 AI, Faculty of Information Technology. GPA: 9.19/10 (3.86/4.0), 130 accumulated credits (official transcript, Sep 2026).

Research & professional experience

AI R&D Lead, Realtime Robotics (RTR), Ho Chi Minh City (Feb 2026 - Present)
Developing visual positioning systems for UAVs: cross-view image matching, model compression for edge deployment, and robustness under domain shift.
Research team member, Team HCMUS_TheFangs, VNUHCM, University of Science (2025 - Present)
Mentored by Prof. Long Tran-Thanh (Warwick) and Prof. The Anh Han (Teesside). Research on LLM agents in game-theoretic settings; international shared-task participation.
Freelance software engineer, Remote (Canada-based client) (2025 - Present)
Full-stack development with React/Next.js and cloud infrastructure; continuous delivery.

Selected honors & awards

Winner, FSE-AIWare 2026 Agentic Python Dependency Resolution Competition, team MEMRES
1st Place, NeurIPS 2025: Mouse vs AI Robust Foraging Competition (both tracks)
1st Place, NakbaArchiveClassifier Shared Task, NakbaNLP @ LREC 2026
1st Place (x2), AbjadNLP Shared Tasks @ EACL 2026
Top 3.3% (138/4,182 teams), NVIDIA Nemotron Model Reasoning Challenge (Kaggle), 2026
Third Prize, Vietnam Student AI Olympiad (National Round), 2025
2nd Prize, Vietnam Student AI Olympiad (Regional Round), 2025
2nd Place, MMLoSo 2025 Shared Task (IJCNLP-AACL Workshop)
2nd Place, ALTA 2025 Shared Task (ADE Normalization)
2nd Place, TRACS @ WASP 2025 (IJCNLP-AACL)
2nd Place (Top 2/172), Dig4Bio Raman Transfer Learning Challenge (Kaggle), 2025

Publications

* equal contribution

Journal Articles

Interface Focus Royal Society journal, accepted 2026

Payoff Scaling Shapes Cooperation in LLM Agents across Languages

Trung-Kiet Huynh, Duy-Minh Dao-Sy, Thanh-Bang Cao, Phong-Hao Le, Hong-Dan Nguyen, Phu-Quy Nguyen-Lam, Minh-Luan Nguyen-Vo, Hong-Phat Pham, Phu-Hoa Pham, Thien-Kim Than, Chi-Nguyen Tran, Huy Tran, Gia-Thoai Tran-Le, Alessio Buscemi, Le Hong Trang, The Anh Han

How the size of the stake and the prompt language change the strategies LLM agents play in cooperation dilemmas. Earlier title: More at Stake.

arXiv PDF

Conference Papers

Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery

NeurIPS 2026 Main Track Co-first author

Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery

Chi-Nguyen Tran*, Duy-Minh Dao-Sy*, Trung-Kiet Huynh*, Phu-Quy Nguyen-Lam, Phu-Hoa Pham, Long Tran-Thanh

SkyPart: a lightweight prototype-based semantic-part head on a DINOv2 ViT that matches drone views to satellite tiles under fog, rain, snow and darkness. State of the art on weather-robust drone-to-satellite geo-localization with the fewest parameters among top methods.

arXiv PDF

MIST: Reliable Streaming Decision Trees for Online Class-Incremental Learning via McDiarmid Bound

NeurIPS 2026 Main Track Co-first author

MIST: Reliable Streaming Decision Trees for Online Class-Incremental Learning via McDiarmid Bound

Phu-Hoa Pham*, Chi-Nguyen Tran*, Phu-Quy Nguyen-Lam, Duy-Minh Dao-Sy, Trung-Kiet Huynh, Long Tran-Thanh

Streaming decision trees that learn new classes online from a data stream, with split decisions backed by a McDiarmid concentration bound and knowledge inherited across splits.

arXiv PDF

Unlocking Compositional Generalization in Continual Few-Shot Learning

NeurIPS 2026 Main Track

Unlocking Compositional Generalization in Continual Few-Shot Learning

Phu-Quy Nguyen-Lam, Phu-Hoa Pham, Duy-Minh Dao-Sy, Chi-Nguyen Tran, Trung-Kiet Huynh, Long Tran-Thanh

Holistic training with a slot-attention router, then compositional inference that scores queries part by part, so a continual few-shot learner generalizes to unseen combinations of known parts.

arXiv PDF

Training-Free Cultural Alignment of Large Language Models via Persona Disagreement

NeurIPS 2026 Main Track

Training-Free Cultural Alignment of Large Language Models via Persona Disagreement

Trung-Kiet Huynh, Duy-Minh Dao-Sy, Tuan Nguyen, Chi-Nguyen Tran, Phu-Hoa Pham, Phu-Quy Nguyen-Lam, The Anh Han, Long Tran-Thanh

Aligns a frozen LLM to the moral preferences of a country without training: country-conditioned personas vote, and only the decisions they agree on are kept.

arXiv PDF

Visual Robustness and Neural Alignment in a Shared Foraging Task: The Mouse vs. AI Benchmark

NeurIPS 2026 Evaluations & Datasets Track

Visual Robustness and Neural Alignment in a Shared Foraging Task: The Mouse vs. AI Benchmark

Marius Schneider, Joe S. Canzano, Yuchen Hou, Jing Peng, A. Deepu, U. Karan, Phu-Hoa Pham, Chi-Nguyen Tran, Duy-Minh Dao-Sy, Phu-Quy Nguyen-Lam, Trung-Kiet Huynh, S. Azeglio, Spencer L. Smith, Michael Beyeler

The benchmark paper of the NeurIPS 2025 Mouse vs. AI competition, co-authored with the organizers as a member of the first-place team.

Report

Competition, Challenge & Shared-Task Papers

HCMUSDroneBoys at SemEval-2026 Task 11: Asymmetric Counterfactual Debiasing and Rank-Sensitive Logical Invariance Adaptation for Syllogistic Reasoning

SemEval 2026 Shared Task Co-first author

HCMUSDroneBoys at SemEval-2026 Task 11: Asymmetric Counterfactual Debiasing and Rank-Sensitive Logical Invariance Adaptation for Syllogistic Reasoning

Chi-Nguyen Tran*, Duy-Minh Dao-Sy*, Trung-Kiet Huynh, Phu-Hoa Pham, Phu-Quy Nguyen-Lam

1st Place (tied), Task 11 Subtask 1, perfect score 100.0

Separates logical form from content so an LLM judges whether a syllogism is valid, not whether it sounds plausible: structure-disentangled prompting, valid-to-invalid counterfactual augmentation, and high-rank QLoRA adaptation.

PDF ACL Anthology Code

HCMUS_RepeatedGames at SemEval-2026 Task 12: CausalRAG: Synergizing Causal Graph Retrieval and Extended LoRA for Abductive Reasoning

SemEval 2026 Shared Task Co-first author

HCMUS_RepeatedGames at SemEval-2026 Task 12: CausalRAG: Synergizing Causal Graph Retrieval and Extended LoRA for Abductive Reasoning

Duy-Minh Dao-Sy*, Chi-Nguyen Tran*, Trung-Kiet Huynh*, Phu-Quy Nguyen-Lam, Phu-Hoa Pham

Tied 5th, Task 12 (score 0.90)

Finds the most plausible cause of a real event from retrieved documents: hybrid BM25 plus dense retrieval for causal keywords, then LoRA on both attention and MLP layers of a 32B model with 0.81% trainable parameters.

PDF ACL Anthology Code

MEMRES: A Memory-Augmented Resolver with Confidence Cascade for Agentic Python Dependency Resolution

FSE 2026 AIWare Competition Co-first author

MEMRES: A Memory-Augmented Resolver with Confidence Cascade for Agentic Python Dependency Resolution

Chi-Nguyen Tran*, Duy-Minh Dao-Sy*, Trung-Kiet Huynh*, Phu-Hoa Pham, Phu-Quy Nguyen-Lam, Vu Nguyen

Winner, FSE-AIWare 2026 Agentic Python Dependency Resolution Competition

An agentic resolver that repairs broken Python environments with a self-evolving memory, a confidence cascade over version candidates, and a build-and-reflect loop.

arXiv ACM DL Code

RoadFury at the ICST 2026 Tool Competition: Self-Driving Car Testing Track

ICST 2026 Tool Competition Co-first author

RoadFury at the ICST 2026 Tool Competition: Self-Driving Car Testing Track

Chi-Nguyen Tran*, Duy-Minh Dao-Sy*, Trung-Kiet Huynh*, Phu-Hoa Pham, Phu-Quy Nguyen-Lam

A test prioritization tool for self-driving-car simulation: a Transformer reads the full geometric sequence of each road, so failure-revealing tests run first. APFD 0.804, above the prior best of 0.781.

IEEE Xplore Code

Early-Stage Prediction of Review Effort in AI-Generated Pull Requests

MSR 2026 Mining Challenge

Early-Stage Prediction of Review Effort in AI-Generated Pull Requests

Duy-Minh Dao-Sy, Trung-Kiet Huynh, Phu-Quy Nguyen-Lam, Phu-Hoa Pham, Chi-Nguyen Tran, Dinh-Ha-Duong Nguyen, Bao-Tran Truong

Predicts at creation time which agent-authored pull requests will cost maintainers the most review effort; flagging the top 20% catches 69% of total effort.

arXiv PDF

JHARNA-MT: A Copy-Augmented Hybrid of LoRA-Tuned NLLB and Lexical SMT with Minimum Bayes Risk Decoding for Low-Resource Indic Languages

MMLoSo 2025 IJCNLP-AACL Workshop Co-first author

JHARNA-MT: A Copy-Augmented Hybrid of LoRA-Tuned NLLB and Lexical SMT with Minimum Bayes Risk Decoding for Low-Resource Indic Languages

Trung-Kiet Huynh*, Duy-Minh Dao-Sy*, Chi-Nguyen Tran*, Phu-Quy Nguyen-Lam, Phu-Hoa Pham, Dinh-Ha-Duong Nguyen, Dien Dinh, Hong-Buu-Long Nguyen

2nd Place, MMLoSo 2025 Language Challenge

Retrieval, LoRA-tuned NLLB and a lexical SMT branch, combined by Minimum Bayes Risk decoding for low-resource Indic translation.

PDF ACL Anthology

Systematic Evaluation of Machine Learning and Transformer-Based Methods for Scientific Telescope Literature Classification

WASP 2025 IJCNLP-AACL Workshop Co-first author

Systematic Evaluation of Machine Learning and Transformer-Based Methods for Scientific Telescope Literature Classification

Trung-Kiet Huynh*, Duy-Minh Dao-Sy*, Chi-Nguyen Tran*, Phu-Quy Nguyen-Lam, Phu-Hoa Pham, Dinh-Ha-Duong Nguyen, Dien Dinh, Hong-Buu-Long Nguyen

2nd Place, TRACS Shared Task

Classical ML against transformer models for classifying telescope-related scientific literature.

PDF ACL Anthology

DRAGON: Dual-Encoder Retrieval with Guided Ontology Reasoning for Medical Normalization

ALTA 2025 Shared Task

DRAGON: Dual-Encoder Retrieval with Guided Ontology Reasoning for Medical Normalization

Duy-Minh Dao-Sy, Phu-Quy Nguyen-Lam, Phu-Hoa Pham, Chi-Nguyen Tran, Trung-Kiet Huynh, Bao-Tran Truong

2nd Place, ALTA 2025 Shared Task

Cascaded dual-encoder retrieval plus ontology reasoning to normalize adverse drug events to MedDRA terms.

PDF ACL Anthology