le van
truong

AI RESEARCHER & ENG

DRAG & FLING

Lê Văn Trường

(Van-Truong Le)

AI Researcher

LLMs, NLP & Fact-Checking

Information Retrieval & RAG

Honors IT Student @ VNU-HCM

VNU-HCMUS

work

Le Van Truong (Van-Truong Le)·SELab & AITHENA Lab (HCMUS) · VNG Saigon AI Hub

buildingsystemsthatreason.findingwhytheyfail.

View Research Selected Papers (7+)Download CV (PDF)

about Scroll to Research Dossier · Honors GPA 9.50/10.0 · ACM ICMR LSC '26 Rank #1 papers

01 · Research Overview·VNU-HCM University of Science (SELab & AITHENA Lab)

I build systems that reason.I research how they fail.

I'm an AI researcher and engineer exploring large language models, fact verification, and retrieval-augmented systems. I turn research ideas into evaluated, working AI systems.

Undergraduate researcher in the Honors Information Technology Program (GPA 9.50/10.0) at VNU-HCM University of Science, conducting research at SELab, AITHENA Lab, and VNG – Saigon AI Hub. First author on 7 research manuscripts across ACM ICMR (LSC '26), KES '26, ICMV '25, FDSE '25, FJCAI '26, and arXiv.

Full Research Program (6 Directions)Selected Papers ↓CV (PDF)

Flagship Research · ACM ICMR LSC '26Rank #1 · Task 3 QA

ChronoSeek: Temporal-First Agentic RAG

Transforms 725,000+ egocentric lifelog images into a Neo4j Event Knowledge Graph and Milvus vector index, orchestrated by a time-bound LangGraph agent.

Task 3 QA

1000 / 1000

Zero-Shot MRR

0.607 (10.8×)

Graph Latency

< 6 ms

Open Paper Webpage Direct PDF

01 · First-Author Papers

6 published/accepted (ACM ICMR LSC, KES, ICMV, FDSE, 2× FJCAI) + 1 arXiv preprint

02 · Honors Program GPA

9.50 / 10

B.Sc. Honors Program in IT (23CNTN), VNU-HCM University of Science (10.0 in Prob & Stats)

03 · ACM ICMR LSC '26 Task 3

Rank #1

Perfect 1000.0 / 1000.0 QA score with ChronoSeek Temporal-First Agentic RAG

04 · Scholarships & Honors

12 + 22

12 major scholarship programs (Odon Vallet, Pony Chung, VietSeeds) & 22 academic/coding honors

Research Thesis & Pillars

Why I Study How Reasoning Systems Fail

Explore All 6 Research Directions

Primary Direction · Pillar 01Fact Verification & Hybrid RAG

Evidence-Grounded Reasoning

Standard dense embeddings compress complex claims into flat vectors that blur crucial factual predicates, dates, and temporal order. My primary research replaces surface similarity with structured evidence grounding: temporal Event Knowledge Graphs (ChronoSeek), retriever-aligned hard negative cross-encoders (SemEval/KES '26), and provenance-aware expert routing (GraphCURE).

Ensemble Bi-Encoders (BiLSTM-Attention over E5) + Retriever-Aligned Cross-Encoder Reranking (KES '26).

10.8× MRR gain over CLIP in temporal lifelog retrieval & 0.9454 S@10 across 10 languages

Inspect Direction

Pillar 02 · Empirical Auditing & Error Taxonomies

Diagnostic Failure Analysis

High benchmark averages frequently hide systematic reasoning collapses. Across 5,922 Vietnamese legal statute questions and 300 expert-graded essays (FJCAI '26), we categorized 7 distinct failure modes—showing that models with strong multiple-choice accuracy still hallucinate abrogated statutes in open-ended legal reasoning.

→ 7 failure modes audited across 5,922 MCQs & 300 essays (Cohen's κ = 0.854)

Pillar 03 · Inductive Biases Under Regime Shifts

Hybrid Analytical + Learned Priors

Domain priors—such as graph Laplacian Fiedler vectors in road disruption estimation (ResiliRoad) or facial skin masks in remote photoplethysmography (SKIA)—accelerate learning when aligned with the signal, but can hurt when violated. I design gated residual architectures that adaptively down-weight analytical priors.

→ 120-regime phase diagram + 13 OpenStreetMap city networks; -42.7% talking rPPG MAE

Pillar 04 · Real-Time & Reproducible Systems

Research-to-Engineering Integration

Research ideas only prove their worth when they survive real-world system constraints: sub-6 ms Neo4j graph queries under a live 4-minute competition clock at ACM ICMR LSC '26, 124 FPS two-stage video anomaly detection, and deterministic Zenodo evaluation harnesses.

→ < 6 ms Neo4j EKG latency on 33,471 events & 124 FPS real-time vision inference

Interactive Research Atlas

How My Research Domains Connect

Explore All 6 Research Directions

Interactive Topology · Drag & Click Any Node

6 Domains · 9 Synergy Links

Active: 01 · Fact Verification

Drag nodes to tug connected domains

Domain 012 Papers · 3 Systems

Automated Fact Verification & Evidence Reranking

Distinguishing semantically similar claims that differ in factual truth across languages and inference regimes.

Core Research Questions

How should negative examples be mined so that a Cross-Encoder reranker learns factual discrimination rather than surface lexical overlap?
Can we improve multi-class fact verification accuracy without updating model parameters by exploiting provenance-aware prediction history and constrained expert routing?

Key Methods

BiLSTM-Attention Token Pooling over Multilingual E5Retriever-Aligned Dense Hard Negative MiningCross-Encoder Reranking (BGE-M3 / PhoRanker)Constrained Expert Routing & Asymmetric NEI Deferral (GraphCURE)

Connected Papers (2)

KES 2026 · Procedia CSPublished

Enhancing Multilingual Fact-Checking Retrieval with Ensemble Bi-Encoders and Architecture-Aware Hard Negative Reranking

FJCAI 2026 · arXiv:2604.16270Accepted

From Benchmarking to Reasoning: A Dual-Aspect, Large-Scale Evaluation of LLMs on Vietnamese Legal Text

Connected Systems (3)

Multilingual Fact-Checking Retrieval Pipeline GraphCURE Training-Free Fact Verification Framework

Full Research Program

02 · Selected Publications

Featured Research Papers

Showing 3 flagship first-author manuscripts with 1-sentence Quick Summaries, direct PDF links, and interactive Paper Pages.

All 8 Papers & Preprints

01PublishedACM ICMR 2026 · LSC '26· ACM Digital Library · Scopus (Rank B)

First Author2026

ChronoSeek: Temporal-First Agentic RAG with Event Knowledge Graph for Lifelog Retrieval

Van-Truong Le, Minh-Triet Tran, Viet-Tham Huynh

The 9th Annual ACM Workshop on the Lifelog Search Challenge (LSC '26), co-located with ACM ICMR 2026, Amsterdam, Netherlands

Quick Summary (What + Result)

Designed a temporal-first Event Knowledge Graph (EKG) schema in Neo4j encoding NEXT, SAME_PLACE, and SIMILAR_ACTIVITY edges over segmented lifelog events.

LSC '26 Task 3 (QA) Score

1000.0 / 1000.0 (Rank #1)

Highest score among all 15 participating teams globally

Zero-Shot Event-Aware MRR

0.607

10.8× improvement over static CLIP (0.056) and SigLIP (0.00)

Oracle Event-Aware MRR

1.000

Validated on both 7K-image dev set and 118K-image 3-month subset

PDF (Direct)Paper Page SystemDOILab Repo

Architecture / Figure

High-level architecture of ChronoSeek showing Offline Indexing and Online Retrieval & Reasoning

Figure 1: High-level architecture of ChronoSeek combining offline Event Knowledge Graph + Milvus indexing with online LangGraph Agentic Orchestration and Fast-Path fallback.

02PublishedKES 2026 · Procedia CS· Elsevier Procedia Computer Science · Scopus (Rank B)

First Author2026

Enhancing Multilingual Fact-Checking Retrieval with Ensemble Bi-Encoders and Architecture-Aware Hard Negative Reranking

Van-Truong Le, Duy-Minh Nguyen-Tran, Tung Le, Huy Tien Nguyen

30th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems (KES 2026), Procedia Computer Science 292 (2026) 1845–1853

Quick Summary (What + Result)

Proposed a two-stage multilingual fact-checking pipeline combining a 5-model BiLSTM-Attention E5-Large ensemble with a BGE-M3 Cross-Encoder reranker.

Cross-Lingual S@10

0.8833

+8.0% over Stage-1 Dense Ensemble (0.8033) & +2.46% above 1st-place PINGAN AI (0.8587)

Overall S@10

0.9163

Monolingual S@10 = 0.9215; reduces mono-to-cross-lingual gap from 9.40% to 3.82%

Negative Mining Ablation

0.9163 vs. 0.2015

Dense-mined hard negatives (0.9163) vastly outperform Random (0.8049) and BM25-mined (0.2015)

PDF (Direct)Paper Page System

Architecture / Figure

Two-stage multilingual fact-checking retrieval architecture with 5x Bi-Encoder Ensemble and BGE-M3 Cross-Encoder

Figure 1: Overview of the two-stage framework: 5× E5-Large + BiLSTM-Attention Ensemble (Stage 1) and Dense Hard-Negative BGE-M3 Reranker (Stage 2).

03PreprintarXiv:2609.11166 · Zenodo· arXiv (cs.LG) · Zenodo Permanent Archive

First Author2026

When does a spectral prior help graph learning? Connectivity-loss estimation under road-network disruptions

Van-Truong Le

arXiv Preprint (cs.LG) & Zenodo Archival Release (v1.0.0)

Quick Summary (What + Result)

Formulated a bounded residual spectral GNN estimator combining closed-form Fiedler edge sensitivity with learned nonlinear corrections for simultaneous multi-edge deletions.

Zero-Shot OSM Spatial MAE

0.096 → 0.057

Residual GCN paired gain of 0.0391 [95% CI: 0.0151, 0.0662] on expanded 13-area panel

Targeted Disruption Gain

+0.0257 (Zero-Shot) / +0.0622 (LOCO)

Statistically supported improvement under edge-betweenness targeted attacks

20,000-Node Screening Speedup

56× Faster (23.17 ms vs. 1,302.4 ms)

Amortized over 1,000 screened disruption scenarios per intact eigensolver setup

PDF (Direct)arXiv Zenodo Paper Page Code System

Architecture / Figure

ResiliRoad workflow diagram showing base spectrum, fast prior, and GNN residual correction

Figure 1: ResiliRoad workflow separating intact spectral prior computation from GNN residual correction.

View Full Publications Archive (8 First-Author Papers + 7 Conference Works) →

03 · Projects & Interactive Systems

Built Systems, Demos & Code

Interactive competition interfaces, open-source repositories, and system architectures across 18 technical projects.

All 18 Projects & Case Studies

ChronoSeek Interactive Lifelog Search Competition Interface

Live Competition Interface · ACM LSC '26

Interactive System Spotlight

ChronoSeek Interactive Retrieval & Fast-Path UI

Built for live 4-minute competition rounds at ACM ICMR LSC '26: combines natural-language LangGraph query decomposition, color-coded Neo4j event timelines, temporal context inspection, and 1-click DRES evaluation submission.

Live QA Score

1000 / 1000 (Rank #1)

Archive Scale

725,949 Images

Inspect System Architecture Paper Page

01Research Systems

Lab System· 2026

ChronoSeek

Temporal-First Agentic RAG with Event Knowledge Graph for Lifelog Retrieval

A temporal-first agentic retrieval system transforming 725,000+ lifelog images into a Neo4j Event Knowledge Graph and Milvus vector store, orchestrated by a time-bound LangGraph agent. Achieved a perfect 1000.0/1000.0 score (Rank #1) on Task 3 (Question Answering) at ACM ICMR LSC 2026.

LSC '26 Task 3 (QA)

1000.0 / 1000.0

Zero-Shot MRR

0.607

PyTorch 2.1LangGraphQwen2.5-VL-7Bgoogle/siglip-base-patch16-224

Explore Case Study

Paper

02Research Systems

Public· 2026

ResiliRoad

Spectral-Prior Residual Graph Learning for Road-Network Disruption Screening

A reproducible graph machine learning framework that combines analytical first-order Fiedler spectral sensitivity priors with residual GNNs (GCN, GraphSAGE, Edge-MPNN) to estimate road-network connectivity loss across 13 OpenStreetMap areas in 6 Asian countries, achieving a 56× screening speedup on 20,000-node networks.

OSM Spatial-Cluster MAE

0.096 → 0.057

Targeted Disruption Gain

+0.0257 (GraphSAGE) · +0.0622 (LOCO GCN)

PyTorch 2.11NetworkX 3.6Residual GCN (9,745 params)Residual GraphSAGE (14,593 params)

Explore Case Study

Paper GitHub

03Research Systems

Public· 2026

Multilingual Fact-Checking Retrieval Pipeline

Ensemble BiLSTM-Attention Bi-Encoders & Dense Hard-Negative Cross-Encoder Reranking

A two-stage multilingual and cross-lingual claim verification retrieval system evaluated on SemEval-2025 Task 7 across 10 languages. Achieves 0.9163 Overall S@10 and 0.8833 Cross-Lingual S@10 (+8.0% gain from reranking), surpassing the competition winner in cross-lingual retrieval.

Cross-Lingual S@10

0.8833

Overall S@10

0.9163

PyTorchSentence-Transformersintfloat/multilingual-e5-large-instruct (+ Custom BiLSTM-Attention)BAAI/bge-reranker-v2-m3 (568M)

Explore Case Study

Paper

Explore All 18 Engineering Systems & Case Studies →

04 · Research Labs & Appointments

Experience

Active research appointments at SELab, AITHENA Lab, and VNG – Saigon AI Hub. Click “Details” on any role to expand full technical bullets.

All 7 Roles & Delegations

01 · Sep 2025 – Present

Undergraduate AI Researcher

SELab — VNU-HCM University of Science

Software Engineering & AI Laboratory · Multimedia Retrieval, Computer Vision, rPPG & Biometric Security

Conducting first-author research on temporal-first agentic lifelog retrieval (ACM ICMR LSC '26), motion-robust physiological computing / rPPG (ICMV '26), and biometric super-resolution privacy leakage evaluation (FDSE '26).

ChronoSeek (ACM ICMR LSC 2026)SKIA / SKIA-Net (ICMV 2026)FLeX (FDSE 2026 · Springer CCIS)

02 · Sep 2025 – Present

Undergraduate AI Researcher

AITHENA Lab — VNU-HCM University of Science

Automated Fact Verification, Natural Language Processing & Large Language Models

Conducting first-author research in multilingual and Vietnamese automated fact-checking retrieval, dense-and-cross-encoder reranking architectures, and diagnostic reasoning evaluation of Large Language Models.

Multilingual Fact-Checking Retrieval (KES 2026 · Procedia CS)Vietnamese Legal LLM Evaluation (FJCAI 2026 · arXiv:2604.16270)ViMELD & Automated Fact-Checking Systems

03 · Nov 2025 – Oct 2026

Research Member

VNG – Saigon AI Hub

Applied AI Lab Batch 1 (VR Story Train AI) · Joint VNG Group & VNU-HCM Initiative

Selected into one of 10 teams in Saigon AI Hub's inaugural Applied AI Lab to research and engineer multimodal agentic systems and VR clinical-communication training environments.

Lumina-HT2 (ACM ICMR 2026)CrisisX (IEEE ISMAR 2026 Poster)ChronoSeek (ACM ICMR LSC 2026)

05 · Scholarships & Honors

Highlighted Scholarships & Awards

Showing 6 representative distinctions out of 34 verified scholarships and competitive awards.

All 34 Scholarships & Awards

Scholarship Recipient2026

Odon Vallet Scholarship

Odon Vallet Foundation & Rencontres du Vietnam

Awarded by the Odon Vallet Foundation (France) and Rencontres du Vietnam to outstanding university students and young researchers across Vietnam demonstrating exceptional academic and scientific research achievements.

ScholarshipsInternational

Scholarship Recipient2025–2026

Pony Chung Foundation Scholarship

Pony Chung Foundation (Republic of Korea)

Prestigious international scholarship awarded by the Pony Chung Foundation of South Korea to top-ranking undergraduates at Vietnam National University Ho Chi Minh City with strong academic excellence, research potential, and leadership record.

ScholarshipsInternational

Scholarship Recipient2025–2026

VPBank Future Banker Scholarship

Vietnam Prosperity Joint Stock Commercial Bank (VPBank)

Merit-based corporate talent scholarship recognizing top technology and computer science undergraduates with strong academic performance, research accomplishments, and community leadership.

ScholarshipsNational

Scholarship Recipient2023–2025

VietSeeds Foundation Scholarship (Multi-Year Fellow)

VietSeeds Foundation

Comprehensive multi-year fellowship providing financial sponsorship, intensive leadership training, mentorship, and career readiness development across the 2023–2024 and 2024–2025 academic cycles.

ScholarshipsNational

Scholarship Recipient2023–2026

100% Tuition Waiver & Academic Encouragement Scholarships (6× Consecutive)

University of Science, VNU-HCM

Awarded across 6 consecutive evaluation cycles (2023–2024, 2024–2025, and 2025–2026) for maintaining top-tier academic GPA (3.73/4.00) and moral conduct scores across the Faculty of Information Technology.

ScholarshipsUniversity / Institutional

Academic Honor2026

Presenter, 3rd HCMUS Young Scientists Conference

University of Science, VNU-HCM

Presented empirical AI research findings at the 3rd Young Scientists Conference hosted by VNU-HCM University of Science, contributing to the AI and Information Systems track.

Research & AcademicUniversity / Institutional

View All 12 Scholarships + 22 Awards & 24 Verified Certifications →

06 · About

About Le Van Truong

Full Biography & Skill Matrix Life & Writing (Vietnamese · VI)

Le Van Truong presenting at VNG – Saigon AI Hub Applied AI Lab Demo Day 2026

Applied AI Lab Demo Day 2026VNG – Saigon AI Hub

“Rise Together: From Research to Real World” — Presenting multimodal AI research & systems in Ho Chi Minh City.

I am an undergraduate AI researcher and engineer in the Honors Program in Information Technology at the University of Science, Vietnam National University Ho Chi Minh City (VNU-HCM), where I maintain a 9.50/10.0 GPA (with 10.0/10.0 in Probability & Statistics and Calculus 2B).

Across my research at SELab (HCMUS — multimedia lifelog retrieval at ACM ICMR LSC '26, rPPG physiological sensing at ICMV '26, and biometric privacy at FDSE '26), AITHENA Lab (HCMUS — automated fact verification at KES '26, NLP, and legal LLM reasoning diagnostics at FJCAI '26), and the VNG – Saigon AI Hub Applied AI Lab, I investigate a unifying question: how can AI systems retrieve, verify, and reason over complex multimodal or structured evidence without succumbing to surface-level shortcuts?

Honors IT Mathematical Foundation (VNU-HCM University of Science)

10.0 / 10

Probability and Statistics

10.0 / 10

Calculus 2B

9.7 / 10

Discrete Mathematics

9.6 / 10

Applied Mathematics and Statistics

9.3 / 10

Linear Algebra

Beyond the Lab · Personal SpaceVietnamese Content (VI)