Specialty
A training-free approach that ranks visual tokens by both current relevance and changing importance performed strongly across several multimodal language-model benchmarks, including tests with 88.9% pruning.
Scientific publication4 min read
A new preprint describes LipDA, a detector that uses mismatches between lip motion and head movement to identify LipSync forgeries and estimate which generator family produced them.
Scientific publication4 min read
An arXiv preprint reports higher within-model caption scores for full fine-tuning, while Video-LLaVA prompts remained below the strongest fine-tuned model and its full fine-tuning and LoRA tests were not completed.
Scientific publication3 min read
A method called MOSAIC reported higher macro-client top-1 accuracy than FedDDA across domain-shift and joint-shift benchmark settings. The study points to a better average across clients rather than a universal win, with only three seeds and no confidence intervals reported.
Scientific publication5 min read
The uncertainty-guided system recorded higher segmentation metrics on an independent cohort, with its largest reported change in tumor core, while a structural tripwire routed 36 of 362 cases to full manual review.
Scientific publication4 min read
A preprint testing FlyVision across computer-vision and biomedical datasets found that it nearly matched ResNet18 on MNIST with far fewer parameters, while larger tasks produced mixed results.
Scientific publication4 min read
A methods preprint describes LiDAR-SAM2, which uses video foundation-model masks to create 4D point labels and train LiDAR-only models without human LiDAR annotation. On SemanticKITTI, the labels produced higher reported scores than SAM2 labels but did not match full ground truth.
Scientific publication4 min read
A preprint reports a video-generation system that combines a unified 3D control interface with streaming output, but its tests also show a modest quality gap and occasional failures in object consistency.
Scientific publication4 min read
An agent trained for video deep research scored higher than its base model on two benchmarks, while an ablation reported fewer interaction turns when tool necessity was weighted.
Scientific publication4 min read
A transformer model for sparse-view computed tomography reported leading image-quality scores in fixed and variable-view tests and was also tested on one real nanoscale mouse-brain pillar.
Scientific publication4 min read
LEGO-OPD combines language-prior and visual-grounding signals and reports higher reasoning averages than evaluated baselines in Qwen3 experiments, with perception gains limited to benchmark scores.
Scientific publication3 min read
A new preprint reports lower simulated time and blood loss for VoxelSage than a deterministic baseline, while stressing that the results do not establish clinical efficacy or safety.
Scientific publication4 min read
A preprint reports a texture-space diffusion system that reconstructs and generates physically based material maps for known 3D objects, with strong synthetic scores but mixed real-data results.
Scientific publication4 min read
An automated framework built from spatial question-and-answer examples reported higher scores for Qwen2.5-VL models across benchmark comparisons, while its method depends on simulator scene metadata.
Scientific publication3 min read
A preprint reports a graph-guided neural system for stitching character motions that adapts transition length, with the best overall comparison scores on LAFAN1-within and the highest naturalness rating in a 20-person study.
Scientific publication4 min read
A preprint describes a computer-vision model that uses top-down maps as spatial memory for navigation and spatial reasoning. It reports strong benchmark results and high success rates in three robot settings, but the evaluations leave broader generalization unanswered.
Scientific publication4 min read
A proposed way to pass knowledge between AI models built around different data types ranked first in seven of eight reported classification settings and most segmentation measures. The gains came with extra training work and a fixed choice about how much teacher information to retain.
Scientific publication4 min read
A computer-vision preprint reports that its COOL framework completes partial object observations with low reported collision rates across benchmark comparisons, while noting limits in detail and generalization.
Scientific publication4 min read
A preprint reports that FLASH, a training-free method that avoids contrastive decoding for large vision-language models, improved several benchmark results and had the lowest reported inference-time multiplier among the compared methods.
Scientific publication4 min read
A preprint describes UniAfford, a shared model for 2D image and 3D point-cloud affordance grounding, and reports leading scores in cross-dataset tests. The benchmark evidence does not show closed-loop robot manipulation success.
Scientific publication4 min read
A modeling preprint finds that motion can score well as the right action while losing the distinctive pattern of the source clip. Tests on animal motion and human-to-robot data compare action-level recognition with source-instance fidelity, a measure of whether source differences reappear in generated outputs.
Scientific publication5 min read
A new benchmark of story-generating AI systems found that orchestrated systems were more reliable overall, while preserving visual details from one story transition to the next was the clearest weakness.
Scientific publication5 min read
A machine-learning model trained to predict both enlarged perivascular-space scores and their locations achieved the strongest test-set result among four methods, while human observers still agreed more closely with one another than with the model.
Scientific publication4 min read
Seedance 2.0 recorded the highest reported OPIS score at 56.01, while scores across eight evaluated systems ranged from 48.65 to 56.01. The benchmark measures observable object preservation from an initial image, not internal memory.
Scientific publication4 min read
A preprint reports that MomADv2, an autonomous-driving model, had lower long-horizon collision and trajectory errors than comparison systems in benchmark evaluations.
Scientific publication4 min read
A methods preprint tested whether video-training categories should be grouped by compatible update directions before adapter training, finding uneven automated gains across two open models and no human study.
Scientific publication4 min read
A method that refines a multimodal model’s internal states during inference reported higher averages than an initial chain-of-thought approach on two open-weight backbones.
Scientific publication3 min read
A computer-vision preprint reports lower wrist-aligned 3D hand-pose errors across three main benchmarks, including clips where hands disappear from view.
Scientific publication4 min read
An arXiv preprint reports that its alignment-first model posted top or near-top scores across several UAV and remote-sensing benchmarks, while broader sensor settings remain untested.
Scientific publication3 min read
The computer-vision model reported lower camera-motion errors than tested baselines and broadly comparable visual quality, but its rankings depend on the evaluation setup.
Scientific publication4 min read
CMGR outperformed three registration methods on right coronary artery motion in XCAT simulations and selected clinical pairs, while whole-volume results were less consistent.
Scientific publication5 min read
A new arXiv preprint reports that a coding-inspired system kept average benchmark performance close to the uncompressed setup while using 25% of the visual-token budget on Qwen3-VL-8B.
Scientific publication3 min read
A systematic review finds that most deep-learning brain MRI reconstruction studies did not compare image-fidelity scores with radiologist assessments on the same data, leaving key safety questions unanswered.
Scientific publication4 min read
CF-YOLO, a YOLOv11-based detector built to refine context and fine detail, scored higher than YOLOv11n on CTDD but delivered mixed results on external NEU-DET data.
Scientific publication4 min read
A tiny two-recording test found that a parameter-free color extrapolator beat copying the last frame and every trained comparator on held-out copper in both directions. The air-to-chamber advantage was individually separated from zero, while the chamber-to-air interval included zero.
Scientific publication5 min read
A video AI method called Token-Budget Distillation retained strong benchmark scores after aggressive visual-token compression, although its training still depended on the uncompressed teacher model.
Scientific publication5 min read
An anatomy-aware system called CheXtriev reported stronger case-retrieval scores than global and local comparison methods on selected chest radiographs. The gains were especially notable for several lower-prevalence findings.
Scientific publication4 min read
EfficientNetB0 correctly classified 97.36% of 303 held-out mango images, with eight errors, in a study that also deployed the model through a public web app. The result is an initial within-dataset benchmark, not evidence that the tool will deliver the same performance across regions or improve agricultural decisions.
Journal of Bangladesh Academy of Sciences, vol. 50, Supplement 1, p. 114, 20265 min read
The fixed-round PTD model reported sharp speed gains and higher scores across several video benchmarks, while leaving disjoint events and multiple matching targets largely unexplored.
Scientific publication4 min read
A preprint describes one OCR system trained across 13 Indic scripts; its reported overall character error rate was 6.9%, compared with 8.6% for monolingual models.
Scientific publication4 min read