Skip to content
View xixiaouab's full-sized avatar
15 followers · 4 following

Achievements

Achievement: Starstruck

Achievements

Achievement: Starstruck

Block or report xixiaouab

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
xixiaouab/README.md

Xi Xiao

Final-year Ph.D. Candidate · Computer Science · University of Alabama at Birmingham

Homepage · Google Scholar · LinkedIn · Email

I study foundation model post-training, multimodal reasoning, and agentic reinforcement learning. My current interests include visual grounding, adaptive reasoning, agent memory, harness engineering, and recursive self-improvement. At UAB, I am co-advised by Tianyang Wang and Min Xu.

Research experience

  • Oak Ridge National Laboratory · Research Scientist Intern — Design and train an AI4Weather foundation model for the Genesis Mission.
  • Amazon AGI Foundations · Applied Scientist Intern — Post-trained a 200B-parameter frontier MLLM for multimodal and agentic reasoning using SFT and reinforcement learning across more than 100 compute nodes.

Selected research

Project Research focus Resources
ReaLVR Grounding latent visual reasoning in visual evidence Repository · Paper · Project
FORGE Adaptive evidence routing for frozen LLM agents Repository · Paper · Project
VIGIL Counterfactual visual alignment for MLLM reasoning Repository · Paper · Project
StructLoRA Structured, task-aware low-rank model adaptation Repository · Paper · Project
Cascaded Semantics Semantic guidance for vision foundation model adaptation Repository · Paper · Project
Prompt Fusion Discovery Layer-specific prompt fusion through differentiable search Repository · Paper · Project

See my homepage for the full publication list and background.

Recognition

  • SC 2025 Best Paper Award and ACM Gordon Bell Prize Finalist (team) — ORBIT-2.
  • CVPR 2026 Outstanding Reviewer and ICML 2026 Silver Reviewer.

Opportunities and contact

I am seeking Summer 2027 research internships in multimodal reasoning, agentic AI, and foundation model post-training.

Email: xxiao [at] uab [dot] edu

Popular repositories Loading

  1. ReaLVR-code ReaLVR-code Public

    ReaLVR: grounding latent visual reasoning in visual evidence. Training and evaluation code.

    Python 50 5

  2. VIGIL VIGIL Public

    VIGIL: counterfactual visual alignment to mitigate visual laziness in multimodal LLMs (ECCV 2026).

    Python 4

  3. Visual-Instance-aware-Prompt-Tuning Visual-Instance-aware-Prompt-Tuning Public

    Visual Instance-aware Prompt Tuning (ACM MM 2025): PyTorch architecture and experiment scaffold.

    Python 3

  4. PROBE PROBE Public

    PROBE (WACV 2026): PyTorch research scaffold for self-supervised visual prompting in cross-domain road damage detection.

    Python 2

  5. StructLoRA StructLoRA Public

    StructLoRA (ACL 2026): structured, task-aware low-rank adaptation. Architecture and core training scaffold.

    HTML 2 1

  6. Prompt-Fusion-Discovery Prompt-Fusion-Discovery Public

    Layer-specific prompt fusion search for adapting frozen vision foundation models (ECCV 2026).

    Python 2