| Apr 2026 | Our paper ReasoningBomb has been accepted to ACM CCS 2026! We propose an RL-based framework that crafts short, natural-language prompts to trap LRMs into pathologically long reasoning, with a constant-time surrogate reward enabling 4.39×10⁵× training speedup. Just 10% malicious traffic cuts benign throughput by 49.8% and monopolizes 64.3% of compute. Check out our paper, website, code, and dataset. |
| Mar 2026 | ROM is now on arXiv. We propose a lightweight streaming framework to detect and mitigate overthinking in large reasoning models at the token level — a compact detection head (<0.01% extra parameters) monitors reasoning in real-time and triggers early exit when redundant steps are detected, achieving 93.51% accuracy with 47.2% shorter responses. Check out our project page, code, and dataset. |
| Feb 2026 | ReasoningBomb is now on arXiv. Check out our website, code, and dataset. |
| Sep 2024 | Passed my Qualifying Exam. |