Kyungmin Park

Kyungmin Park, Taesup Kim

Safety-aligned Large Language Models (LLMs) remain vulnerable to interventions during inference that redirect generation toward harmful outputs. Recent work attributes this to shallow safety, where alignment concentrates in the first few output tokens. We show that shallow safety is a special case of a broader inference-time vulnerability, in which short token injections at any generation step can substantially alter subsequent safety behavior. We also find that a model's alignment with refusal directions in its hidden states does not predict its robustness to such injection, revealing that internal state alone does not determine generation behavior under perturbation. To address this, we align models directly on generation trajectories constructed by simulating mid-sequence perturbation, and show that this improves robustness to mid-sequence injection and generalizes to attacks that exploit early-token generation. Our work argues that robust safety alignment requires training on the generation process itself, not only its outputs.

Abhirami Harilal, Kyungmin Park, Michael Andrews et al.

The online Data Quality Monitoring system (DQM) of the CMS electromagnetic calorimeter (ECAL) is a crucial operational tool that allows ECAL experts to quickly identify, localize, and diagnose a broad range of detector issues that would otherwise hinder physics-quality data taking. Although the existing ECAL DQM system has been continuously updated to respond to new problems, it remains one step behind newer and unforeseen issues. Using unsupervised deep learning, a real-time autoencoder-based anomaly detection system is developed that is able to detect ECAL anomalies unseen in past data. After accounting for spatial variations in the response of the ECAL and the temporal evolution of anomalies, the new system is able to efficiently detect anomalies while maintaining an estimated false discovery rate between $10^{-2}$ to $10^{-4}$, beating existing benchmarks by about two orders of magnitude. The real-world performance of the system is validated using anomalies found in 2018 and 2022 LHC collision data. Additionally, first results from deploying the autoencoder-based system in the CMS online DQM workflow for the ECAL barrel during Run 3 of the LHC are presented, showing its promising performance in detecting obscure issues that could have been missed in the existing DQM system.

Davendra Maharaj, Tu Pham, Peter Meiring et al.

Dynamic GNN inference has exhibited effectiveness in High Energy Physics (HEP) experiments at High Luminosity Large Hadron Collider (HL-LHC) due to strong capability to model complex particle interactions in collision events. Future HEP experiments will involve detectors that produce 10x more collision data to help unlocking physics discoveries. Due to limitations in offline compute capacity and storage, revamped trigger systems require FPGAs to run ultra-low-latency Machine Learning models for online filtering of useful events with low power consumption. State-of-the-art GNN accelerators relied on static graph structures, but this assumption breaks down in real-time HL-LHC trigger systems and edge-based dynamic GNN models where edge embeddings change in-place based on neighbor node embeddings at runtime. We propose DGNNFlow, a novel dataflow architecture for real-time edge-based dynamic GNN inference applications, especially HL-LHC trigger systems, with three key contributions. First, we introduce hardware support for dynamic computation of edge embeddings. Second, we resolve data dependencies in edge-based dynamic GNN dataflow, where edge embedding is formulated using its source and target nodes. Third, we perform input dynamic graph construction auxiliary setup for complete support of models without pre-defined edge embeddings. We deployed DGNNFlow using AMD Alveo U50 FPGA to evaluate end-to-end latency on-board at 200 MHz clock frequency. DGNNFlow achieved 1.6x-6.3x speedup and 0.22x power consumption compared to GPU (NVIDIA RTX A6000) with batch sizes from 1 to 4, 3.2x-5.1x speedup and 0.25x power consumption compared to CPU (Intel Xeon Gold 6226R). Our complete implementation is publicly available on GitHub.