#Research
Curated coverage on «Research»: releases, research, benchmarks, open-source.
Physicists Find a Way to Instantly Move an Entangled Quantum System From Chaos to Order
An international team describes in Nature Physics a method that instantly moves an entangled quantum system from a chaotic state to a fully controlled one.
ByteDance Researchers Find a Weak Spot in DeepSeek's Memory-Saving Approach
A paper from ByteDance with co-authors from Princeton, Stanford and Berkeley questions the accuracy of DeepSeek models built on a memory-saving technique.
Humanoids Learn to Serve in Tennis Like Pros: Planner and Tracker Split Apart
How humanoids mimic the style of professional tennis players: a hierarchical architecture with separate planner and tracker, adaptive speed control to bridge the…
APEX Compresses Diffusion Latents 128x and Beats a 256-Token Model on ImageNet-512
SWIM saves compute: audio-video generation gains up to 30.7% over Best-of-N
SAIL Keeps Its Target in Frame Using Audio and Text, and Beats Rivals on Two Benchmarks
Wearable Stereo Camera Estimates Food Volume With 9.1% Error, No Reference Object Needed
Researchers Compare CNN, ViT and Mamba Attention Against Human Gaze
Three architectures, three domains, three metrics: eye tracking shows CNNs look broadly, ViTs lock onto micro-features, and Mamba is strong in medicine but diverges…
Stroop Test Shows VLMs Read the Text in an Image Before They See the Color
Researchers ran contrastive and generative VLMs through the classic Stroop test. When text and color conflict, the models more often answer with the written word.
MemReg Outperforms Pairwise Methods and SLAM in LiDAR Point Cloud Registration
MIRAGE anchors video-based 3D reconstruction to LiDAR and beats COLMAP on three datasets
OMNI: One Compact Encoder Absorbs the Knowledge of 10 Models for Pathology
TerraAgent checks whether the AI actually looks at the image: hallucinations drop from 17.1% to 9.8%
TimesFM 3.0 Cuts University Building Energy Forecast Error by 43%
Pretrained TimesFM 3.0, with no local fine-tuning, cut day-ahead MAE for a university building by 43% versus a naive baseline. XGBoost stays close behind.
Wasserstein BaryBooster Turns Diffusion Realism into Event Forecast Accuracy
The authors recast event-sequence forecasting as conditional generation and show that no method wins on both accuracy and realism. BaryBooster offers a way out.