arXiv:2608.22975v1 Announce Type: new Abstract: Embodied multimodal agents must answer from growing observation streams under a fixed per-decision token budget. We formalize this constraint through f...
HGSQ is a real‑time aerial small‑object detector that uses a heatmap‑guided sparse query strategy to focus computation on foreground regions. It introduces a lightweight Heatmap Budget Predictor to generate a foreground budget map, and then employs Heatmap‑Guided Sparse Query Selection, Heatmap‑Gated Lite Snake Convolution, and Adaptive Query‑Decoder Budgeting to efficiently process only small‑object areas. On NWPU VHR‑10 and VisDrone2019, HGSQ achieves 95.10 mAP50 and 54.8 mAP50 respectively while running at 96 FPS with only 48.6 GFLOPs on an RTX 4070.
The paper introduces ReadKV, a query‑adaptive quantization method for KV caches that stores keys and values in progressive codes, allowing each decoding query to allocate different precision prefixes for keys and values. By optimizing a calibrated distortion objective under fixed budgets, ReadKV achieves near‑optimal allocation and demonstrates that query‑dependent access can outperform any query‑independent reader at the same read budget. Experiments on six base models show that reading only four bits on average from an eight‑bit cache increases perplexity by at most 0.66% while reducing logical reads and retained capacity compared to a 16‑bit cache, and it achieves lower latency than existing codecs on an NVIDIA A10G.
Autonomous industrial inspection requires strictly deterministic real-time inference under tight memory and computational constraints. However, existing Video Instance Segmentation (VIS) models suffer...
RT-DETRv4 introduces a cost‑effective distillation framework that leverages Vision Foundation Models to improve lightweight real‑time object detectors. The method uses a Deep Semantic Injector to embed high‑level VFM representations into deep detector layers and a Gradient‑guided Adaptive Modulation strategy to adjust semantic transfer intensity based on gradient norms. These techniques yield consistent performance gains across DETR‑based models without adding inference overhead, achieving state‑of‑the‑art COCO AP scores at high FPS.
arXiv:2611.07994v1 Announce Type: new Abstract: Tiny object detection remains challenging because tiny instances occupy only a few pixels, and their fine-grained details are easily lost during featur...
Joint Subspace Sparse Attention (JSSA) is a new method that reduces KV‑cache traffic in long‑context LLM decoding by selecting tokens at the token level while preserving query–key structure. It learns a shared low‑dimensional query–key subspace through a tractable spectral solution and implements a gather‑free indexed‑attention kernel that supports per‑KV‑head selections. Across various attention architectures, JSSA maintains near full‑KV accuracy, avoids generation‑length inflation, and can achieve up to 6.43× decoding throughput on a single GPU for 128K context, even at 700B‑parameter scale.
arXiv:2608. 03918v1 Announce Type: cross Abstract: Efficient long-video understanding requires vision--language models (VLMs) to reason over a small number of frames selected as sparse visual evidence.
arXiv:2606. 00144v1 Announce Type: cross Abstract: Speculative decoding speeds up autoregressive decoding by using a drafter to propose multiple tokens that a verifier validates in parallel.
Sparse attention reduces the KV-cache traffic that dominates long-context LLM decoding, but practical methods largely operate at coarse granularity. Token-level selection offers finer control, yet pre...
DetM-Tracker is a lightweight Video Instance Segmentation framework designed for autonomous industrial inspection. It reduces memory and computation by condensing instance features into four-scalar descriptors, dynamically allocating query capacity, and using entropy-gated skip connections to focus processing on uncertain regions. The method achieves roughly 20× lower inference latency and 6× less VRAM usage than dense-feature baselines.
The paper investigates an output‑selection bottleneck in frozen DETR‑family segmentation models, showing that many useful mask proposals are computed but never exposed. A lightweight selector called HYDRA, trained only on cached outputs, can recover up to +7.41 mIoU on ADE20k and COCO and +9.4 class‑macro prompt‑IoU on SAM 3 across eight domains by selectively choosing better candidates. The study demonstrates that evaluating segmenters should consider both exposed masks and the hidden candidates they suppress.