Daniel Shiu-Hong Kao

PhD student @cvml_nus | Prev intern @MSFTResearch,
MPhil & BSc @HKUST

[Daniel Shiu-Hong Kao @danielshkao]

Thrilled to announce that our newest accepted paper in #ICLR2026 has been released and open-sourced.

Project page: https://t.co/6F3TBKEfWd
Code: https://t.co/k3Ks7fic7q

[Daniel Shiu-Hong Kao @danielshkao]

I won't be making it to #ICLR2026 because of a final exam conflict. 🥲 Our posters for CoT-RVS and CoT-Seg will still be available.

Sad to miss the networking and vibes, but duty calls! Good luck to everyone presenting—catch you all at the next conference!

[Daniel Shiu-Hong Kao @danielshkao]

We will present CoT-RVS in the ICLR 2026 main conference and CoT-Seg in the RSI workshop😁 See you then!

[Daniel Shiu-Hong Kao @danielshkao]

CoT-Seg is accepted to the ICLR2026 workshop on AI with Recursive Self-Improvement! 🎉

🔗Project page: https://t.co/iqMxtHTxhN

TL; DR. A training-free framework for reasoning segmentation with self-evaluation and -refinement.

[Daniel Shiu-Hong Kao @danielshkao]

@danxuhk Thanks Prof. Xu!

[Daniel Shiu-Hong Kao @danielshkao]

Thrilled to announce that our newest accepted paper in #ICLR2026 has been released and open-sourced.

Project page: https://t.co/6F3TBKEfWd
Code: https://t.co/k3Ks7fic7q

[Daniel Shiu-Hong Kao @danielshkao]

@keeplea58276507 Thanks!

[Zhengzhong Tu @_vztu]

Dear NeurIPS reviewers, please be reminded to delete the GPT prompts next time :)

[Dawn Song @dawnsongtweets]

🚀 Really excited to launch #AgentX competition hosted by @BerkeleyRDI @UCBerkeley alongside our LLM Agents MOOC series (a global community of 22k+ learners & growing fast). Whether you're building the next disruptive AI startup or pushing the research frontier, AgentX is your launchpad. Two tracks:
- Entrepreneurship: Build agent-powered products & startups
- Research: Explore the frontiers of LLM Agents technology
📅 Registration opens TODAY! Submissions due end of May
🏆 Winners showcase at our Agents Summit to industry leaders and VCs in August @UCBerkeley! 🌟

🙏 Tremendous thanks to our incredible sponsors @Amazon @huggingface @LambdaAPI @MistralAI @Google @GroqInc @schmidtsciences; proud to partner w. leading VCs in the space @Accel @BainCapVC @BessemerVP @lightspeedvp @MayfieldFund @NEA! Stay tuned—more sponsors/partners AND exciting prizes/credits/resources info will be announced soon! 🚀

⏰ Register now at https://t.co/1tXZOB2BVL and join us in shaping the future of AI! #AgentX #AI

[Zhenjun Zhao @zhenjun_zhao]

StreamGS: Online Generalizable Gaussian Splatting Reconstruction for Unposed Image Streams

Yang LI, Jinglu Wang, Lei Chu, Xiao Li, Shiu-hong Kao, Ying-Cong Chen, Yan Lu

tl;dr: DUSt3R->initial coarse reconstruction; 2D matches refine this reconstruction

https://t.co/labG8Fjem1

[Daniel Shiu-Hong Kao @danielshkao]

It generates a chain of thoughts (CoT) to process highly implicit queries.

[Daniel Shiu-Hong Kao @danielshkao]

We just released a new preprint:
"Think Before You Segment: High-Quality Reasoning Segmentation with GPT Chain of Thoughts", the first CoT-Reasoning Segmentation model.

Check out the project page at https://t.co/Z7W8ziEXrR.

[Daniel Shiu-Hong Kao @danielshkao]

It also supports easy image-based controls like casual scribbles, bounding boxes, or points.

[Chia Hong Hsu @ch_dinocat35]

Excited to share my first research project—Invertible Guided Consistency Training (iGCT).

iGCT enables fast, deterministic forward & backward mapping along the PFODE path without relying on a diffusion model in training—a first step toward integrating inversion & guidance into Consistency Training.

Website: https://t.co/SuKpq90lhW
Paper: https://t.co/Zxk86AkoLR

[AK @_akhaliq]

Open O1: A Model Matching Proprietary Power with Open-Source Innovation

[Daniel Shiu-Hong Kao @danielshkao]

We have released our new work UVRM, where we introduce a pose-free training and testing method for scalable 3D reconstruction model from monocular videos!

https://t.co/oPX9RlND3D

[Daniel Shiu-Hong Kao @danielshkao]

Thrilled to share that I've received the prestigious "Stars of Tomorrow" Internship Award of Excellence from @MSFTResearch Asia! I'm extremely grateful for the opportunity to work with such an innovative and talented team 🫶🏼.

[Jindong Gu @gu_jindong_]

Call for interns in Google!

Are you interested in building Multi-modal and Reliable Agents with Gemini?

I am looking for talented interns and student researchers to join our team at Google.

If interested, please fill out this form https://t.co/IpegZSVqCD or PM me.

[AI at Meta @AIatMeta]

New from Meta FAIR — Byte Latent Transformer: Patches Scale Better Than Tokens introduces BLT, which for the first time, matches tokenization-based LLM performance at scale with significant improvements in inference efficiency & robustness.

Paper ➡️ https://t.co/0iamZCRnMN

[Daniel Shiu-Hong Kao @danielshkao]

Today is my last day as an intern at Microsoft Research Asia. I am incredibly grateful for the opportunity to learn and grow alongside my mentor Dr. Xiao Li and talented colleagues. Thank you all for your support, guidance, and inspiration throughout this journey.