Skip to main content
Search

Xidian University

6 related articles

Subscribe to updates
  1. arXiv

    Depth boundaries are a sliver of pixels but nearly half the error: E3Stereo cuts IGEV boundary error from 3.77 to 3.23 pixels

    Stereo matching error concentrates at depth boundaries: on IGEV, geometric edge pixels are a small fraction of evaluated pixels yet contribute about half of total absolute disparity error. E3Stereo learns geometric edges from disparity supervision and uses the same edge map both to guide full-resolution disparity residual correction and to weight boundary errors during training, jointly optimized with the stereo backbone. On Scene Flow it reduces boundary end-point error by 13.4%, 14.3% and 9.5% for RAFT-Stereo, IGEV and MonSter while lowering overall error, but direct transfer to real datasets is worse than IGEV and only target-domain fine-tuning brings gains across all reported ETH3D and KITTI metrics.

  2. Single-cell 3D genome boundaries survive sparse data: cross-cell alignment lets boundary strength split an embryo subpopulation RNA could not

    Single-cell 3D genome data are so sparse that domain boundaries are hard to call reliably, and positional disagreement across cells blocks comparison of boundary frequency and strength. CellRoB couples single-cell boundary detection with cross-cell alignment that minimizes an insulation-score loss, improving boundary accuracy and robustness to contact dropout on mouse brain Dip-C and human GM12878 scMicro-C data from 40 kb down to 5 kb. On multimodal mouse embryo data, boundary-strength profiles split the EPI cells that the RNA embedding left as one group.

  3. Co-frequency signal separation at 235K parameters: complex attention cuts collapse and keeps phase

    Single-channel blind separation of co-frequency communication signals has relied on real-valued networks or attention modules blind to complex structure, which cannot exploit the phase coupling between the I and Q rails. ComplexLightweightSepNet embeds a complex squeeze-and-excitation block in a complex convolutional encoder-decoder, scaling complex features with a real-valued channel weight that preserves phase, at 235,525 parameters. On 5 Hz co-frequency synthetic mixtures across four modulations it reaches a five-seed mean SDR of 2.31 dB with 0.62 dB standard deviation; paired Wilcoxon tests do not reach alpha = 0.05, so the result is positioned as a working point at the 235K scale, not an inferred advantage.

  4. arXiv

    LT-OPD uses on-policy self-distillation on its own generated trajectories to lift average retained performance at 5% visual tokens from 68.6% to 82.3%

    The work proposes LT-OPD, which supervises a student keeping only 5% of visual tokens on its own generated trajectories using a frozen full-token copy of the same model as teacher, together with a curriculum that progressively lowers the token budget, raising average retained performance on nine benchmarks with Qwen3.5-4B from 68.6% to 82.3% while cutting KV cache by 85.2% and prefill FLOPs by 85.4% with no added inference overhead.

  5. arXiv

    VA-Adapter lets an ultrasound foundation model guide the echo probe, cutting trained parameters about 33-fold while lowering guidance error below strong baselines

    The work proposes a Vision-Action Adapter (VA-Adapter) inserted into the deep layers of a frozen ultrasound foundation model image encoder (EchoCLIP, USFM, BiomedCLIP) to online-inject understanding of individual 3D cardiac structure by encoding historical vision-action sequences; on a dataset of 178 adults, 356 expert scans and 1.31M image-action pairs, it reaches lower translation and rotation mean absolute error than strong probe-guidance baselines with roughly 2.61M-3.97M trainable parameters.

  6. Bioinformatics

    Regulatory-prior-guided attention preserves biological structure during unpaired single-cell RNA-ATAC integration

    The work presents scHPGT, a single-cell Heterogeneous Prior-Guided Transformer that integrates unpaired RNA and chromatin accessibility profiles in a shared latent space using modality-specific encoders, a prior-guided cross-modal Transformer that constrains gene-peak attention with regulatory links, and a domain-adversarial objective, improving clustering agreement, label transfer and biological structure preservation across PBMC3k, mouse spleen, CITE-seq/ASAP-seq PBMC and PBMC10k benchmarks while avoiding forced correspondence of unmatched or condition-specific states in partial-overlap and condition-shift settings, with attention-derived links recovering regulatory relationships, marker-gene regulatory regions, transcription factor programs and regulatory activity profiles.

You’re all caught up