Research field
Artificial intelligence, computing, robotics, energy, and engineering.
A replay-free method for continual learning scored higher across the reported benchmarks and showed no measured retention loss on TRACE, though its routing has not been tested under severe distribution shifts.
Scientific publication4 min read
A shared humanoid policy reported higher success than LHM-Humanoid on seen and unseen object-transport tasks, but success fell as the number of objects increased.
Scientific publication4 min read
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 preprint describes a robot that watches for meaningful changes after human-object interactions and uses a vision-language model to plan assistance without a task request. The reported comparison is described as comparable, but the available extraction omits the numerical results and uncertainty estimates needed to judge how close the systems were.
Scientific publication4 min read
A preprint reports valid-and-unique molecular yields and more distinct elite architectures, while a digit-image control favored Soft on distribution matching.
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
A preprint reports favorable shifts on several gender-bias measures across three language models, while general-capability averages remained close to the corresponding base models.
Scientific publication5 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
A preprint describes TACD, a distillation method that limits terminal supervision and reports stronger benchmark results, faster generation and lower peak GPU memory across text-to-motion tests.
Scientific publication4 min read
A preprint reports that PI-SCAN identified and characterized overlapping radio interference in simulated data, but its ability to detect every source fell as mixtures grew more crowded.
Scientific publication4 min read
A modeled power grid reached 91.50% accuracy on the first 400 MNIST test images, while results varied with network changes, noise and the way images were represented.
Scientific publication4 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
CrashSim generated safety-critical scenarios that tracked real-world crash behavior and crash-type statistics more closely than competing methods. In closed-loop tests, the resulting nuCrash set produced much wider performance differences among five autonomous-driving planners.
Scientific publication3 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
The tested methods reached more than 99% of the study's HVmax on several graph sets, while the preprint says they may not recover every best trade-off.
Scientific publication4 min read
A reinforcement-learning retriever called MO-IKE reported higher combined scores than DR-IKE on Llama and Mistral tests, and recorded an overall score of 24.5 on UniEdit.
Scientific publication4 min read
A review of public discussions, interviews and repositories finds that many software specifications mix requirements with implementation detail, while practitioners use the workflow selectively.
Scientific publication4 min read
A preprint study of software repositories found that coding-agent adoption coincided with more activity, but repositories with more structured committed AI practices had smaller increases in some static-analysis quality measures. The authors stress that the findings are associations, not proof that configuration caused the difference.
Scientific publication4 min read
In a laboratory test with one human collaborator and a seven-degree-of-freedom robot, an adaptive controller recorded a 95% average success rate, compared with 83% for a fixed-stiffness version and 69% for a deterministic baseline.
Scientific publication4 min read
Thirteen frontier models were tested on 20 computational-biology tasks, and none scored above 0.48 overall; performance was lower on very large inputs and deep analysis chains.
Scientific publication4 min read
A theoretical preprint finds that delayed cache misses preserve key online bounds, while offline victim choice and unequal fetch costs remain harder.
Scientific publication4 min read
A methods study found higher reproducibility for claim-locked language-model reports in fMRI tests, while separate audits showed why stability is not the same as correctness.
Scientific publication4 min read
A modeling study found that one speech-act-based forecasting design led on three dialogue datasets and in low-data tests, while transfers between datasets remained uneven.
Scientific publication4 min read
A preprint proposes a 41-part vocabulary for construction activities and demonstrates four selected physical actions on robot hardware.
Scientific publication3 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
A preprint reports that cached Event Tokens produced the best reported ranking score and higher retrieval scores across several industrial recommendation comparisons.
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 theoretical preprint proposes a way to compress weighted graphs and trained neural networks with an explicit error budget, but its demonstration covers only three post-training cases.
Scientific publication4 min read
A preprint benchmark reports that an episode-based AI memory system scored higher than tested full-context and retrieval methods while using much less retrieved text.
Scientific publication4 min read
A mathematical preprint gives deterministic bounds linking a code’s excess quadratic discrepancy to how unevenly its correction-radius balls cover a finite Hamming space. The result does not show closeness to any particular perfect code or establish that all admissible perfect codes exist.
Scientific publication4 min read
A constructed example cuts constrained risk in half, but the classical ordering returns when the model class is unrestricted.
Scientific publication5 min read
LM-X recorded higher average task success than the tested controls while exposing three online signals for progress, next events and local action reliability.
Scientific publication3 min read
TacForcing updates unfinished action blocks with execution-time tactile feedback and reported the highest average success rates in both simulated and real-world evaluations. The study was limited to six tasks in each setting and two robot platforms.
Scientific publication4 min read
A paper-derived training pipeline produced a 60 billion-token corpus and was linked to higher writing scores, while reasoning results stayed close to direct fine-tuning.
Scientific publication4 min read
Six vision-language models showed more agreement when ranking 3D objects along affective, or feeling-related, adjective pairs than along unrelated pairs, but less than on geometric controls. The pattern varied sharply across object categories and changed when researchers altered the text used to define each axis.
Scientific publication5 min read
A preprint study of graph-based question answering found that carefully controlled rewrites damaged performance mainly during subgraph construction, not entity linking or final decoding.
Scientific publication4 min read
EVEREST recorded the strongest overall mask-overlap and pixel-level precision-recall scores on SocioSeg, while the study does not show whether the result generalizes beyond that benchmark.
Scientific publication4 min read
DCGC reported the best average score in a six-benchmark comparison, while tests of shuffled drafts suggested that the original draft still mattered.
Scientific publication4 min read