Weichen Li (@weich_li) profile picture

PhD student @UChicago. Research Intern @Microsoft.

Chicago, IL
Joined March 2024
Photo by Weichen Li (@weich_li)Photo by Weichen Li (@weich_li)
Weichen Li retweeted
Excited to share that I’ll be presenting our #ICML2026 paper in Seoul: ExVerus: Verus Proof Repair via Counterexample Reasoning 📍 Poster: Thu, Jul 9, 5:00–6:45 PM KST, Hall A #2004 💡 DL4C Workshop: Fri, Jul 10, Session 2, 3:20–4:50 PM KST, Hall A (Board 100-119) Paper: openreview.net/forum?id=FNDk… Happy to discuss! This work focuses on LLM-based Verus proof repair. Verus is a practical Rust-based deductive verification framework that has been used to verify large-scale critical systems, such as cluster management controllers (Anvil) and virtual machine security modules (VeriSMo). In this work, we propose a new paradigm for LLM-based proof repair: repairing a proof via synthesizing, validating, and blocking counterexamples. 1/n
1
7
15
1,164
Thrilled to share our #ICML2026 paper, SemRep : Generative Code Representation Learning with Code Transformations! I will present it on Wednesday (5:00 PM – 6:45 PM HALL A #4014) and also Friday (DL4C Workshop). Happy to Chat! What makes a good code representation? 🤔 Not just for code generation, but also for code reasoning, editing, agents, and all applications where code is the key modality. We identify 4 critical requirements and propose SemRep, a new training paradigm with Generative Code Representation Learning. 🧵1/n
1
3
1
8
1,811
Takeaway: Executable source code explored through semantics-preserving transformations is a better code representation. By making representation learning explicit and generative, we unlock better reasoning, verification, and generalization across all code-centric applications. 📄 Paper: arxiv.org/pdf/2603.13640 💻 Code: github.com/cirrus-uchicago/S… 📊 Slides: bit.ly/4p1XIAA 🧵8/n
1
1
73
Takeaway: Executable source code explored through semantics-preserving transformations is a better code representation. By making representation learning explicit and generative, we unlock better reasoning, verification, and generalization across all code-centric applications. 📄 Paper: arxiv.org/pdf/2603.13640 💻 Code: github.com/cirrus-uchicago/S… 📊 Slides: bit.ly/4p1XIAA 🧵8/n
1
1
106
Weichen Li retweeted
Directed fuzzing is often used to test patches, verify bugs, and reproduce crashes. When using LLM for fuzzing... Everyone: "Let's use LLMs for input-level guidance." 😴 Locus: "Hold my beer. Let's generalize the guidance to the entire program!" 🍺 Locus is not a fuzzer. Locus rewrites the program under test to improve all fuzzers. Check our paper at #icse2026: arxiv.org/pdf/2508.21302. (1/6)
1
5
7
857
Weichen Li retweeted
Drop by our poster #2408 on Dec 3 (Wed) 11– 2 PST at Exhibit Hall C,D,E at #NeurIPS2025 to check our paper "Synthesizing Performance Constraints for Evaluating and Improving Code Efficiency" 🎉📍 We introduce WEDGE: a framework that helps LLMs generate "performance-stressing" tests to expose inefficient code. 🧵👇
8
2
2
19
3,272
🚨 EditLord: New #ICML paper on code editing! Software maintenance accounts for 60% of the software lifecycle, and much of it involves code editing. ⚡️ Developers optimize code efficiency 📖 Reverse engineers refactor decompiled mess 🛡️ Security teams mitigate vulnerabilities 🧵
2
1
1
11
2,593
I’ll present EditLord at #ICML2025 and the co-located #PRAL workshop in Vancouver🇨🇦. Feel free to stop by if you have any questions or want to chat! 🧵9/n
1
2
140