Skip to content
View ElysiaTT's full-sized avatar
๐Ÿ 
Working from home
๐Ÿ 
Working from home

Highlights

  • Pro

Block or report ElysiaTT

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please donโ€™t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this userโ€™s behavior. Learn more about reporting abuse.

Report abuse
ElysiaTT/README.md

Hi there ๐Ÿ‘‹, Iโ€™m ElysiaTT

๐ŸŽ“ CS Undergraduate @ HUST | ๐Ÿ’ผ AI Group Leader @ Unique Studio (่”ๅˆ›ๅ›ข้˜Ÿ)

Typing SVG

  • ๐Ÿ” Research Interests:
    • 3D & Vision: 3D Gaussian Splatting (3DGS), NeRF, and Long Video Generation.
    • Agent & Embodied AI: Vision-Language-Action (VLA) models and intelligent Agent workflows.
  • ๐Ÿ› ๏ธ Tech Philosophy: Bridging the gap between rigorous, "orthodox" 3D computer vision algorithms and the dynamic reasoning of LLM Agents.

๐Ÿ”ฌ Research & Projects

Paper2Rebuttal | Evidence-Centric Multi-Agent System for Academic Rebuttals

  • ๐Ÿ† Accepted by ACL 2026 (Main Conference) (Core Contributor)
  • Innovation: Reconstructed rebuttal generation as an evidence-centric planning task. Features atomic review decomposition, high-fidelity context construction, and autonomous external literature retrieval.
  • Impact: Effectively eliminates hallucinations in LLM responses by anchoring every argument to verifiable internal/external evidence.

OmniPresent Unified Framework & Benchmark for Academic Multimodal Agents

  • Designed a system to generate multimedia artifacts (website, posters, videos, PPTs) from research PDFs.
  • Developed benchmark and evaluation metric for papers.

3D-VLA & World Models for Embodied AI

  • Framework Integration: Investigating the deep integration of 3D-VLA architectures with World Models to tackle perception and decision-making challenges in long-tail autonomous driving and embodied scenarios.
  • Latent Space Alignment: Addressing the weak temporal generalization of pure 3D rendering algorithms (e.g., 3DGS) in dynamic environments by exploring optimization strategies to align 3D Gaussian features with the latent space of World Models.

Long-form Video Generation

  • Developing data pipelines with Qwen3/Molmo2 and SAM 3 for temporal consistency and entity tracking.
  • Time-controllable Video Generation Model (In Development).

๐Ÿ›  Tech Stack Matrix

  • AI & Research: PyTorch, 3DGS, LLM Agents, VLA Architectures.
  • Engineering: Python, C/C++, Git, AI-driven development (Cursor/Vibe Coding).

๐Ÿ† Honors & Experience

  • ACL 2026 Main Conference Paper (Accepted, Core Contributor).
  • AI Group Leader @ HUST Unique Studio (2024 - 2026)
    • Organized team technical sharing sessions and hackathon participation.
  • Honor Prize @ National University of Singapore (NUS) Summer Program (2024).
  • Technology Innovation Scholarship @ HUST (2024 - 2025, 2025-2026).

๐Ÿ“ฎ Contact & Links

github contribution grid snake animation

Pinned Loading

  1. HustSoftware_homework HustSoftware_homework Public

    based on Android studio

    Java 2

  2. HUST_graphics HUST_graphics Public

    ๅŽไธญ็ง‘ๆŠ€ๅคงๅญฆ่ฎก็ฎ—ๆœบๅญฆ้™ขๅ›พๅฝขๅญฆไธ“ไธš่ฏพ็จ‹

    Makefile 1

  3. UniqueAI2024SummerCamp UniqueAI2024SummerCamp Public

    Forked from Whale-Dolphin/UniqueAI2024SummerCamp

    Repo of hustunique AI group 2024 summer camp. You can also learn some AI knowledge if you want.

    Python

  4. HUST_machine_learning HUST_machine_learning Public

    Python

  5. isaacsim-teleop isaacsim-teleop Public

    XR-based teleoperation framework using NVIDIA Isaac Sim

    Python 1

  6. codexchange codexchange Public

    Switch Codex accounts without logging out of ChatGPT. A Windows PowerShell tool for managing Codex auth profiles safely.

    PowerShell