arXiv subjects

Ben Stoffelen

Publications and source records attributed to Ben Stoffelen.

3 recordsLinked to original sources

arXiv2026

KernelArc: A Multi-Agent Framework for GPU Kernel Optimization

We present KernelArc, a multi-agent framework for autonomous GPU kernel optimization across heterogeneous workloads. Strategy-specialized agents run in parallel and coordinate through conclusions-only shared memory, a deterministic benchmark guard, and read-only cross-agent state with plateau-triggered drafting. We evaluate KernelArc on NVIDIA H100 and B200 GPUs using category-representative SOL-ExecBench workloads. The resulting implementations span custom BF16 GEMM, static cuBLASLt Expert-API configuration tables, fused mixture-of-experts backward, shape-gated decoder-layer fusion, native NVFP4 grouped-query attention, and paged prefill attention. In the public SOL-ExecBench leaderboard snapshot recorded on August~20, 2026, KernelArc ranked first on every representative L1, L2, Quantization, and FlashInfer task evaluated. The trajectories support the paper's central motivation: shared multi-agent search can broaden exploration and reach stronger incumbents within a fixed candidate budget, while the value of individual coordination features depends on the kernel and optimization stage.

cs.AI↗

arXiv2026

Addressing Detail Bottlenecks in Latent Diffusion for RGB-to-SWIR Image Translation

Latent diffusion models (LDMs) enable efficient image-to-image translation but discard fine spatial details during compression, degrading downstream perception tasks. We identify two bottlenecks: the autoencoder, which loses spatial information, and the conditioning pathway, which further degrades the source signal through naive downsampling. We propose two lightweight, backbone-agnostic fixes: a Source-Conditioned Autoencoder (SCAE) that injects high-resolution source features into the decoder via skip connections, and a Learnable Guidance Encoder (LGE) that replaces naive downsampling with a learned conditioning signal. Evaluated on RGB-to-SWIR translation for driving scenes with two denoiser backbones (U-Net and DiT), our approach improves detection mAP by up to 2x over the latent diffusion baseline, with up to 3.4x gains on small objects (COCO-small, <32^2 px^2), while achieving state-of-the-art FID. We further show that FID and detection performance are poorly correlated, motivating multi-axis evaluation. Results generalise zero-shot to the public RASMD benchmark. We will publicly release test data with annotations, all checkpoints, and training code.

cs.CV↗

arXiv2025

Increasing the Diversity in RGB-to-Thermal Image Translation for Automotive Applications

Thermal imaging in Advanced Driver Assistance Systems (ADAS) improves road safety with superior perception in low-light and harsh weather conditions compared to traditional RGB cameras. However, research in this area faces challenges due to limited dataset availability and poor representation in driving simulators. RGB-to-thermal image translation offers a potential solution, but existing methods focus on one-to-one mappings. We propose a one-to-many mapping using a multi-modal translation framework enhanced with our Component-aware Adaptive Instance Normalization (CoAdaIN). Unlike the original AdaIN, which applies styles globally, CoAdaIN adapts styles to different image components individually. The result, as we show, is more realistic and diverse thermal image translations. This is the accepted author manuscript of the paper published in IEEE Sensors Conference 2024. The final published version is available at 10.1109/SENSORS60989.2024.10785056.

cs.CV↗