Follow the latest research
alphaXiv connects papers, researchers, and organizations, grounding its answers in the underlying work.
OpenAIA smooth, compactly forced three-dimensional flow can develop unbounded velocity in finite time while retaining uniformly bounded kinetic energy.
A coordinated multiagent search produced a proposed resolution and Lean formalization, giving researchers artifacts to inspect, verify, and build upon.
UMass Amherst
UC Berkeley
A brief visual calibration lets robot world models simulate actions in unseen camera views and embodiments without parameter updates or additional training.
Researchers to follow
View all
Yoshua Bengio
President and Scientific Director @ LawZero, Founder and Scientific Advisor @ Mila - Quebec Artificial Intelligence Institute, Canada CIFAR AI Chair @ CIFAR, Full Professor, CS @ Université de Montréal
Andrew Ng
Managing Partner @ AI Aspire, Managing General Partner @ AI Fund, Founder @ DeepLearning.AI, Adjunct Professor, CS @ Stanford University, Chairman and Co-Founder @ Coursera
Are you a researcher? Find your profile
SJTU
Shanghai Innovation InstituteA single open-source model can generate, clone, edit, enhance, and separate speech through natural-language instructions and optional audio context.
NUS
TsinghuaA modular research stack enables controlled comparisons of world–action architectures, training recipes, and data mixtures across simulations and physical robots.
Tencent
ZJUIn this work, we present Gander, an end-to-end model that unifies omni perception, realtime interaction, and agentic capabilities within a single framework. In contrast to turn-based conventional paradigms, Gander continuously receives streaming inputs across multiple modalities, including video, speech, and text, enabling natural full-duplex interaction in both everyday conversations and complex workflow-oriented agent scenarios. Users can interrupt the model at any time, while the model can also proactively provide intermediate feedback or ask follow up questions. To natively support these capabilities, Gander adopts two key architectural designs: 1) It employs a Cerebellum-Brain collaborative framework, in which the Cerebellum is responsible for realtime interaction and omni conversational capabilities, while the Brain handles complex reasoning and higher-level agentic tasks. The two components interact continuously through tool calling and the agent orchestration runtime. 2) The Cerebellum is built upon a streaming Thinker-Talker architecture, user inputs and model outputs are further flattened into an ordered token stream at the chunk level, providing a unified representation for low latency, continuous interaction. We conduct comprehensive evaluations of Gander across four dimensions: conversational ability, omni understanding, interactive capability, and agentic intelligence. Internal human evaluations demonstrate that Gander maintains the natural and expressive spoken dialogue capabilities of SOTA open source models while achieving competitive performance in omni interaction. Gander also demonstrates robustness in challenging real-world scenarios, including background noise interference, multi-party interactions, and backchannel communication. We release Gander together with its models, code, and data to facilitate further research and development in the community.
KAIST
Microsoft
Weak teachers can accelerate stronger models’ verifier-based training while letting them surpass the teachers and retain their own reasoning strategies.
GoogleEditable procedural graphs help language-model agents choose valid next steps, avoid repetitive tool loops, and improve execution plans from trajectory feedback.
YaleA research idea may be novel, coherent, and scientifically plausible, yet its proposed method may remain insufficiently specified for faithful implementation. We study the codification readiness of implementation-facing research-method specifications, defined by whether they provide sufficient methodological information for a competent implementer or coding agent to construct the intended method without unsupported assumptions. We construct evidence-grounded specifications and their supported resolutions from papers, codebases, issue threads, and reproduction artifacts. We introduce IdeaAMBIG, a benchmark of 660 evidence-grounded instances: 163 real-world gaps from reproducibility reports and GitHub issues, and 497 controlled synthetic gaps injected into codification-ready references. IdeaAMBIG evaluates three capabilities: codification-readiness assessment, defect localization, and clarification action generation. Defect localization receives only the specification, whereas clarification additionally receives the annotated defect. Across 13 LLMs, the best model achieves 9.6% Macro Defect Recovery Rate on real-world instances but 80.6% Macro Clarification Action Success Rate when given the defect. In an oracle study, supplying the gold resolution raises the downstream codification-ready rate from 14% to 98%. Across all evaluated models, defect localization is the main bottleneck, with stronger clarification given the defect.
CMU3D queries let feed-forward models maintain dense 3D point trajectories across hundreds of video frames, even through occlusion and field-of-view changes.
Segment-level rewards from a single reference trajectory let language models learn from partial progress on difficult reasoning tasks where outcome rewards provide no signal.
Researchers to follow
View allLinxi "Jim" Fan
Director & Distinguished Research Scientist @ NVIDIA, Previously CS PhD Student @ Stanford University
Terence Tao
Distinguished Professor @ UCLA, Previously Mathematics PhD Student @ Princeton University
Stanford
Microsoft Research
A single pretrained renderer handles complex scenes with mixed geometry, materials, environment lighting, caustics, and volumetric scattering without per-scene training.
Routing harnesses turn deployed agent trajectories and capability feedback into targeted training data, enabling models to improve on interactive tasks.
Researchers can train frontier-scale language and diffusion models with customizable synchronous or asynchronous post-training workflows across NVIDIA and AMD hardware.
Princeton
CornellAn automated theory-discovery loop using simulated behavior produced decision-making theories that predicted held-out human choices better than canonical models.
UCLAThe taxonomy broadens robot safety beyond collisions to include semantic hazards, misalignment, misuse, privacy, and risks spanning development through deployment.
ETHZMulti-view waypoint predictions triangulate into executable 3D trajectories, helping manipulation policies transfer to unfamiliar tasks, objects, and robot embodiments.
Cambridge
MITRobots can adapt long-horizon manipulation plans to changing object states, recover from execution errors, and verify progress without task-specific training.
The paper envisions interfaces generated on demand from user goals and context, enabling bespoke interactions that connect information across traditionally separate applications.
PKU
BAAITactile feedback and future contact prediction help dexterous robots perform fine-grained manipulation while generalizing across clutter, lighting changes, and unfamiliar objects.
