Yifan Zhang
@CoderDoge1108
Followers
344
Following
121
Media
7
Statuses
32
Ph.D. in CS @VanderbiltU | Ex-Intern: SR @GoogleAI, AI @IBMResearch, GenAI @intel, RS @tiktok_us | Ex-MLE @JD_Corporate | AI4Code | DCAI for Systems
Nashville
Joined May 2022
RT @Kexin_Pei: We're excited to invite you to attend DLSP 2025, co-located with IEEE S&P on May 15! Themes:. Secu….
0
6
0
RT @minimario1729: 📢 Excited to share our new paper: Challenges and Paths Towards AI for SWE. We discuss:.🛠️ 6 sub-tasks needed for SWE.🤖 9….
0
36
0
RT @papers_anon: Enhancing Code LLM Training with Programmer Attention. Introduces an eye-tracking path augmentation method to expand progr….
0
2
0
A prime example of how merging formal verification with a disaggregated design can redefine distributed file systems.
🚀 Day 5 of #OpenSourceWeek: 3FS, Thruster for All DeepSeek Data Access. Fire-Flyer File System (3FS) - a parallel file system that utilizes the full bandwidth of modern SSDs and RDMA networks. ⚡ 6.6 TiB/s aggregate read throughput in a 180-node cluster.⚡ 3.66 TiB/min.
0
0
0
This work pushes AI-driven industrial automation forward, bridging LLMs with real-world Cyber-Physical Systems (CPS). 🤖. It highlights how AI enhances scheduling in high-stakes environments like semiconductor fabrication. 🏭✨.
1
0
1
Key results in construction scheduling 📊. 🔹 More accurate gap filling.🔹 Stronger dependency tracking.🔹 Faster schedule adjustments. These gains show the impact of LLM-driven scheduling in real-world construction. ⚡🏗️. 🧵 6/n
3
0
2
Compared to GPT-4o, CONSTRUCTA achieves:.📈 +42.3% in missing value prediction.🔗 +79.1% in dependency analysis.📅 +28.9% in automated planning. These improvements highlight the potential of LLM-driven automation in industrial AI. 🧵 5/n
2
0
2
Expert-aligned scheduling with CPA-DPO 🤖🏗️. CONSTRUCTA uses CPA-DPO, a specialized RLHF method, to align LLM outputs with expert decisions. This enhances task relevance, dependency reasoning, and overall scheduling efficiency. ⏳✅. 🧵 4/n
2
0
2
CONSTRUCTA integrates three key innovations:.✅ Static RAG → Encodes construction-specific knowledge.✅ Contextual RAG → Retrieves real-time task-relevant information.✅ CPA-DPO (RLHF) → Aligns scheduling decisions with expert preferences. 🧵 3/n
1
0
2
Construction scheduling in semiconductor fabrication is complex:.🔹 Dynamic dependencies.🔹 Real-time updates.🔹 Expert-driven decision-making. Traditional methods struggle to adapt. We introduce an LLM-powered scheduling framework to tackle these issues efficiently. 🧵 2/n.
1
0
3
🚀 Excited to share my latest research has been accepted to the proceedings of #NAACL2025!. Developed during my GenAI internship at @intel, this work applies LLMs to automate scheduling and reasoning in construction.🏗️🔧. 📜 ArXiv: 🧵1/n
25
60
497
Excited to serve as the Web Chair of HumanAISE@FSE'25. Paper Deadline: Feb 25, 2025.Website: Join us in shaping the future of Human-Centered AI4SE!.
humanai4se.github.io
HumanAISE Workshop (Co-located with FSE'25 and ISSTA'25 at Trondheim, Norway).
We will host the 1st workshop on human-centered AI for SE, colocated with FSE 2025! Please consider participating and discussing on the exciting topics with us!
0
0
1
Feel free to join our research presentation at 2pm and poster session at 3:30pm🐶 (p.s. the poster has been displayed, but will stay unattended until 3:30pm lol).
My student Yifan Zhang @CoderDoge1108 will present our paper "EyeTrans: Merging Human and Machine Attention for Neural Code Summarization" @FSEconf today 2pm at Pitomba! Very first work merging human attention via eye tracking with Transformer Attention! Artifacts available :)
0
0
6
Photoshop is the new Eclipse😜(confirmed).
Me: “Wasn’t it cloudy when we took our group eclipse pic?” My student: “Ya! I photoshopped the sun in our picture!” 😂😂😂
0
0
3
RT @nick_konz: Excited to announce our new paper in IEEE Trans. in Med. Imaging, "SWSSL: Sliding window-based self-supervised learning for….
github.com
SWSSL - Sliding window-based self-supervised learning for anomaly detection in high-resolution images (IEEE Trans. on Medical Imaging 2023) - mazurowski-lab/swssl-for-anomaly-detection
0
2
0
RT @YuHuang_yh: At least one opening for our teaching track! I'm on the search committee, pls let me know if you have any questions :).
0
3
0
RT @InteNSEWorkshop: First on our agenda, ASTRO: An AST-Assisted Approach for Generalizable Neural Clone Detection. A nice collaboration wi….
0
3
0
I'm thrilled to announce that I will be joining InfraLab at TikTok to conduct AI for system research this summer. See you in San Jose!
1
0
8
My first in person conference at @ASE_conf Doctoral Symposium. Nice to see the beautiful autumn in Detroit🍂
0
1
10
RT @calvinyluo: Excited to share with everyone an accessible, intuitive tutorial on diffusion models! If you're curious about the math behi….
0
312
0