Project AURIONAI Use-Cases Research & Innovation

R&D Topics

Research topics from the taxonomy, each graded for four tiers that form one pipeline: frontier labs, Indonesian AI labs, institution units such as ITB, and organisations such as SEMAI delivering proof of technology, concept and value.

  • 152topics
  • 608grades approved
  • 0grades proposed
The four tiers of the R&D pipeline, 152 R&D topics. Each bar: the topics graded feasible, stretch and not feasible for that tier.152R&D topicsTier A: Frontier AI labs. 146 feasible, 5 stretch, 1 not feasibleFrontier AI labs146Tier B: Indonesian AI labs. 92 feasible, 45 stretch, 15 not feasibleIndonesian AI labs924515Tier C: Institutions such as ITB. 108 feasible, 38 stretch, 6 not feasibleInstitutions such as ITB10838Tier D: Delivery such as SEMAI. 106 feasible, 28 stretch, 18 not feasibleDelivery such as SEMAI1062818feasiblestretchnot feasible

Tier Grades

Tiers: a) frontier AI labs, b) Indonesian AI labs, c) institution units such as ITB, d) organisations such as SEMAI delivering proof of technology, concept and value.

Tier AFrontier AI labs
146feasible
  • feasiblefeasible146
  • stretchstretch5
  • not feasiblenot feasible1
Tier BIndonesian AI labs
92feasible
  • feasiblefeasible92
  • stretchstretch45
  • not feasiblenot feasible15
Tier CInstitution units such as ITB
108feasible
  • feasiblefeasible108
  • stretchstretch38
  • not feasiblenot feasible6
Tier DOrganisations such as SEMAI (PoT, PoC, PoV)
106feasible
  • feasiblefeasible106
  • stretchstretch28
  • not feasiblenot feasible18

Topics

feasiblefeasiblestretchstretchnot feasiblenot feasible
TopicTier ATier BTier CTier DModelsObserved actors
Open-Weight Frontier Releases as Research Substratefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible279OpenAINVIDIAMistral AIAlibaba (Qwen)
Small Language Models (SLM)feasiblefeasiblefeasiblefeasiblestretchstretchfeasiblefeasible156Microsoft AIHugging FaceAlibaba (Qwen)OpenAINVIDIA
Frontier Pretraining Scalingfeasiblefeasiblenot feasiblenot feasiblenot feasiblenot feasiblenot feasiblenot feasible133SpaceXAI (xAI)DeepSeekZ.ai (Zhipu)OpenAIGoogle DeepMind+4
Text-to-Image Foundation Models (Diffusion and Flow Transformers)feasiblefeasiblenot feasiblenot feasiblestretchstretchfeasiblefeasible131Alibaba (Qwen)Black Forest LabsStability AIGoogle
Retrieval-Augmented Generation (RAG)feasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible130AcademiaNVIDIAIntegrators
Real-Time Object Detection (YOLO and DETR Lines)feasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible112BaiduPeking UniversityRoboflowUltralyticsTsinghua University+2
Image Classification Backbone Architectures (CNN and Vision Transformer)feasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible105GoogleMetaMicrosoft AI
Visual Instruction Tuningfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible105Microsoft AIShanghai AI LabHugging FaceGoogleSalesforce+3
Efficiency Architectures (MoE, Sparse Attention, Long Context)feasiblefeasiblestretchstretchfeasiblefeasiblestretchstretch104DeepSeekAcademic groupsOpenAI
Text-to-Video Foundation Modelsfeasiblefeasiblenot feasiblenot feasiblestretchstretchfeasiblefeasible104Alibaba (Qwen)Google DeepMindKuaishouTencent
Text Embeddings for Semantic Searchfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible96Alibaba (Qwen)BAAIGoogleJina AINVIDIA+3
Regional or National Foundation Model Training (Continued Pretraining at 7B to 70B)feasiblefeasiblefeasiblefeasiblestretchstretchnot feasiblenot feasible95GoTo Gojek TokopediaIndosat Ooredoo HutchisonAI Singapore
Efficient Vision Networks for Mobile and Edge Devicesfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible87GoogleMegviiApple
Low-Resource and Regional Language Resourcesfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch83Indonesian NLP communityAI Singapore
Reasoning via Large-Scale RLfeasiblefeasiblestretchstretchnot feasiblenot feasiblenot feasiblenot feasible74OpenAISpaceXAI (xAI)DeepSeekMiniMaxMistral AI
Code Language Modelsfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible65Alibaba (Qwen)Mistral AIOpenAIMetaCognition (owns Windsurf)+3
Semantic, Instance and Panoptic Segmentation Architecturesfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible65MetaNVIDIAGoogle
Domain Adaptation (Vertical Models)feasiblefeasiblefeasiblefeasiblestretchstretchfeasiblefeasible62BloombergAcademic evaluators
Time-Series Foundation Modelsfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible62Amazon
Agents and Tool Usefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible57Frontier labsNVIDIAAcademic benchmark teamsAnthropic
Large-Scale Zero-Shot Text-to-Speechfeasiblefeasiblefeasiblefeasiblestretchstretchfeasiblefeasible53Microsoft AIFunAudioLLMElevenLabsMiniMax+5
Quantization and Edge Inferencefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible50Academic groupsOpenAI
Image and Video Restoration and Super-Resolutionfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible49TencentETH ZurichByteDanceNanyang Technological University
Self-Supervised Visual Representation Learningfeasiblefeasiblenot feasiblenot feasiblestretchstretchfeasiblefeasible46Meta
Guardrails and Safety Classifiersfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible43MetaNVIDIA
Vision-Language-Action Robot Policiesfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible43Stanford UniversityUC BerkeleyPhysical IntelligenceNVIDIAGoogle DeepMind+7
Contrastive Image-Text Encoders (CLIP and SigLIP Lines)feasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible42Google
Video Action Recognition and Temporal Understandingfeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible42MetaGoogleThe Chinese University of Hong Kong
Agentic Software Engineeringfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible41Alibaba (Qwen)OpenAIMistral AICognition (owns Windsurf)Anysphere (Cursor)+3
LiDAR and Multi-Sensor 3D Perceptionfeasiblefeasiblestretchstretchfeasiblefeasiblestretchstretch41University of Texas at AustinnuTonomyMassachusetts Institute of TechnologyStanford University
Scene and Document Text Detection and Recognition (OCR)feasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible41BaiduHuazhong University of Science and TechnologyJaided AIMicrosoft AI
TinyML on Microcontrollersfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible41STMicroelectronicsMLCommonsArmSilicon LabsMIT Han Lab+3
Human Pose Estimationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible40University of SydneyJD.comOpenMMLabCarnegie Mellon UniversityGoogle
Text-to-Music and Song Generationfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible39Suno, Inc.Uncharted Labs (Udio)Google DeepMindMetaMultimodal Art Projection (m-a-p)+5
Massively Multilingual Speech Recognitionfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible38OpenAIMetaNVIDIAAlibaba (Qwen)Carnegie Mellon University+2
Pretrained Bidirectional Text Encodersfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible35AI SingaporeLiquid AIMicrosoft AIAnswer.AI+3
Text-to-Text Pretraining and Summarisationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible32GoogleMetaAllen AICarnegie Mellon University+3
Deep Learning Architectures for Time-Series Forecasting and Classificationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible31IBMTsinghua UniversityGoogleIntelAmazon+3
Generative Image Modelling Methods (GANs, Diffusion and Image Translation)feasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch31LMU MunichRunwayNVIDIAUC Berkeley
World Models (Generative Interactive Environments)feasiblefeasiblenot feasiblenot feasiblenot feasiblenot feasiblenot feasiblenot feasible30Google DeepMindWorld Labs
Face Detection, Recognition and Livenessfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible29InsightFace (Deep Insight)Google
Medical Image Segmentation and Screening AI (Radiology)feasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch29Bo Wang Lab, TorontoDKFZ (German Cancer Research Center)Helmholtz ImagingQure.aiUniversitas Gadjah Mada (UGM)+1
Neural Rerankers for Retrievalfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible29Alibaba (Qwen)BAAICohereSentence TransformersJina AI+3
Open-Vocabulary Object Detection and Groundingfeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible28IDEA Research (International Digital Economy Academy)TencentGoogleTsinghua University
Audio Language Models for Speech Understandingfeasiblefeasiblestretchstretchnot feasiblenot feasiblefeasiblefeasible27Mistral AIMoonshot AINVIDIAAlibaba (Qwen)IBM+2
Document Parsing Vision-Language Modelsfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible27BaiduDeepSeekOpenDataLabAllen AI
Image-to-3D and Text-to-3D Asset Generationfeasiblefeasiblenot feasiblenot feasiblestretchstretchstretchstretch27TencentMicrosoft AIStability AIVAST
Instruction-Based Image Editing and Inpaintingfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible27UC BerkeleyBlack Forest LabsSamsungSkolkovo Institute of Science and TechnologyStepFun
Document, OCR and Chart Understandingfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible26GoogleOpenBMB (Tsinghua NLP Lab and ModelBest)Microsoft AIZ.ai (Zhipu)Mistral AI+3
Industrial Visual Anomaly and Defect Detectionstretchstretchfeasiblefeasiblefeasiblefeasiblefeasiblefeasible26AmazonCEA (French Alternative Energies and Atomic Energy Commission)MVTec Software GmbH
Open Post-Training Recipesfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible25Allen AIGoogleBigScience WorkshopHugging Face+3
Earth Observation Foundation Modelsfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible24IBMNASAGoogle DeepMindAnt GroupAerospace Information Research Institute, Chinese Academy of Sciences (AIRCAS)+4
Real-Time Spoken Dialogue and Low-Latency Voice Generationfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible24KyutaiOpenAIGoogle DeepMindAmazonSesame AI Inc.+3
Data Curation and Synthetic Datafeasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch23Hugging FaceMicrosoft AIStanford-area groups
Machine-Learning Medium-Range Weather Forecastingfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible23Google DeepMindHuawei (Huawei Cloud)Fudan UniversityShanghai AI LabMicrosoft AI+4
Monocular and Stereo Depth Estimationfeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible23The University of Hong KongByteDanceAppleIntelNVIDIA
Multi-Object Tracking and Re-Identificationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible23Huazhong University of Science and TechnologyUniversity of SurreyQueen Mary University of LondonCarnegie Mellon UniversityShanghai AI Lab+1
Streaming and Low-Latency Speech Recognitionfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible23NVIDIAUseful SensorsMistral AIHugging FaceSYSTRAN
World Models (Self-Supervised Video, Robotics)feasiblefeasiblestretchstretchstretchstretchstretchstretch23MetaNVIDIA
Visual Document Retrievalfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible22Alibaba (Qwen)CohereJina AINVIDIAIlluin Technology+3
Anomaly Detection and Predictive Maintenance on Sensor Datafeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible21NECTsinghua UniversityCarnegie Mellon UniversityIBMMLCommons+3
Tabular Foundation Modelsfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible21ELLIS TübingenAcademic group
Click-Through-Rate Prediction and Sequential Recommendationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible20GoogleUniversity of Science and Technology of China (USTC)Alibaba (Qwen)Alibaba GroupNational University of Singapore+3
Omni-Modal and Unified Multimodal Modelsfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible20DeepSeekAlibaba (Qwen)XiaomiOpenBMB (Tsinghua NLP Lab and ModelBest)Amazon+3
Video Understanding with Language Modelsfeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible19Alibaba DAMOKuaishouAllen AIHugging FaceTwelveLabs+3
Deep Learning and Gradient Boosting for Tabular Datafeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible18YandexMicrosoft AIAmazonDMLC communityINRIA+1
Plantation, Crop and Wildlife Monitoring Visionnot feasiblenot feasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible18Gunadarma UniversityUPN Veteran Jawa TimurMultimedia UniversityGoogle
Pretrained Video Analytics for People, Vehicles, Retail and Safetystretchstretchfeasiblefeasiblefeasiblefeasiblefeasiblefeasible18IntelNVIDIAQualcomm
Distillation of Frontier Models into Smaller Onesfeasiblefeasiblestretchstretchstretchstretchnot feasiblenot feasible17Alibaba (Qwen)DeepSeek
Optical Flow and Local Feature Matchingfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible17Princeton UniversityETH ZurichMagic LeapZhejiang UniversityUniversity of Freiburg
Graph Neural Networks and Graph Transformersfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible16Stanford UniversityMicrosoft AIUniversity of AmsterdamMilaUniversity of Cambridge+3
Keyword Spotting and Audio Event Detection on Microcontrollersfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible16ArmSilicon LabsMLCommonsSTMicroelectronicsGoogle+2
Controllable and Personalised Image Generationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible15Stanford UniversityTencent
Document Layout, Table Structure and Key Information Extractionfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible15Microsoft AIBaiduNaver
Image Matting and Background Removalstretchstretchfeasiblefeasiblefeasiblefeasiblefeasiblefeasible15Nankai UniversityBria AIUniversity of AlbertaUniversity of WashingtonByteDance
Unified Multimodal Models for Understanding and Generationfeasiblefeasiblenot feasiblenot feasiblestretchstretchstretchstretch15ByteDanceBAAISalesforceTencent
3D Human Body Models, Mesh Recovery and Motion Generationfeasiblefeasiblestretchstretchfeasiblefeasiblestretchstretch14UC BerkeleyTencent
Audio-Driven Talking Heads and Lip-Syncfeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible14ByteDanceXi'an Jiaotong UniversityTencentNVIDIA
Massively Multilingual Machine Translationfeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible14MetaGoogleUniversity of Helsinki (Helsinki-NLP)AI4Bharat (IIT Madras)
Promptable Segmentation Foundation Models (Segment Anything)feasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible14MetaKyung Hee UniversityBo Wang Lab, Toronto
Efficient Acoustic Encoder Architectures for Speech Recognitionfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch13GoogleNVIDIAk2-fsaUC BerkeleyBaidu
Neural Vocodersfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible13Kakao EnterpriseNVIDIAGoogle DeepMindDescriptNaver+1
Universal Machine-Learned Interatomic Potentialsfeasiblefeasiblestretchstretchfeasiblefeasiblestretchstretch13Microsoft AIMetaEPFLHarvard UniversityOrbital Materials+5
Video Editing and Controllable Video Generationfeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible13Alibaba (Qwen)DecartNetflixShanghai AI LabThe Chinese University of Hong Kong+1
Biomolecular Structure Prediction and Co-Foldingfeasiblefeasiblestretchstretchfeasiblefeasiblestretchstretch12Google DeepMindIsomorphic LabsBoltz PBCMIT (Jameel Clinic)ByteDance+3
Computational Pathology Foundation Modelsfeasiblefeasiblenot feasiblenot feasiblestretchstretchstretchstretch12Mahmood Lab, HarvardPaigeMicrosoft AIProvidence
Medical Vision-Language Modelsfeasiblefeasiblestretchstretchfeasiblefeasiblestretchstretch12GoogleMicrosoft AIGoogle DeepMindStanford UniversityPaige
Small-Data Reasoning Fine-Tuning and Distillationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible12Stanford UniversityUniversity of WashingtonAllen AIContextual AIUniversities and nonprofits
Training-Data Licensing and Content Provenance for Generative Mediafeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible12Adobe Inc.Bria AI
Few-Step Distillation of Diffusion Modelsfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible11ByteDanceTsinghua UniversityStability AIBlack Forest Labs
LLM-Based Machine Translationfeasiblefeasiblefeasiblefeasiblestretchstretchfeasiblefeasible11ByteDanceTencentUnbabelGoogle
Multi-Camera Bird's-Eye-View 3D Perceptionfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch11Nanjing UniversityShanghai AI LabPhiGent RoboticsUniversity of Texas at AustinStanford University+2
Multilingual and Regional Text Encodersfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible11MetaJohns Hopkins University (CLSP)GoogleINRIAFlax Community (Hugging Face community event)
Protein Language Models and Generative Protein Designfeasiblefeasiblenot feasiblenot feasiblestretchstretchnot feasiblenot feasible11EvolutionaryScaleMetaMicrosoft AIBoltz PBCChai Discovery
Self-Supervised Speech Representation Learningfeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible11MetaMicrosoft AI
Speech Enhancement and Audio Source Separationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible10SpeechBrain communityAlibaba (Qwen)MetaINRIA
Fraud and Outlier Detection on Transaction Datafeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible9Carnegie Mellon UniversityNortheastern UniversityFeaturespaceGoogleHSBC+2
General Audio and Music Representation Learningfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible9MIT (Jameel Clinic)XiaomiMultimodal Art Projection (m-a-p)GoogleTencent+1
Joint Audio and Video Generationfeasiblefeasiblenot feasiblenot feasiblestretchstretchfeasiblefeasible9Character.AILightricksOpenMOSSKuaishou
Probabilistic Ensemble Weather Forecasting with Machine Learningfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible9Google DeepMindGoogleECMWFNVIDIAINRIA+1
Real-Time Road, Lane and LiDAR Segmentation on Embedded Hardwarestretchstretchfeasiblefeasiblefeasiblefeasiblefeasiblefeasible9Huazhong University of Science and TechnologyKU LeuvenHalmstad UniversityVolvo GroupIntel
Speaker Diarizationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible9pyannoteAINVIDIABrno University of Technology (BUT Speech@FIT)OpenMOSS
Compact Text Encoders Through Distillationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible8Hugging FaceGoogleMicrosoft AI
Deep Reinforcement Learning Algorithms for Controlfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible8OpenAIUC BerkeleyGoogle DeepMindMcGill UniversityGoogle
Driving World Models and Closed-Loop Simulationfeasiblefeasiblenot feasiblenot feasiblestretchstretchstretchstretch8WayveWaymoTeslaNVIDIAHuawei (Huawei Cloud)
Generative Recommendationfeasiblefeasiblestretchstretchstretchstretchstretchstretch8KuaishouMeituanMeta
Genomic DNA Language Modelsfeasiblefeasiblestretchstretchfeasiblefeasiblestretchstretch8Arc InstituteNVIDIAUniversity of Southern CaliforniaPrime IntellectGoogle DeepMind+1
GUI and Computer-Use Agentsfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible8Z.ai (Zhipu)Microsoft AIAnthropicGoogle DeepMindTsinghua University+1
On-Device Text-to-Speechfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible8hexgradAmpixaRhasspyNeuphonicSilero+1
Reference-Driven Human Image Animationfeasiblefeasiblestretchstretchfeasiblefeasiblestretchstretch8ByteDanceAlibaba (Qwen)University of TrentoSnap Inc.
Expressive and Controllable Speech Synthesisfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible7Columbia UniversityMaya ResearchAmphion (open-mmlab/Amphion)NVIDIANari Labs+1
Feed-Forward 3D Reconstruction and Visual Geometryfeasiblefeasiblestretchstretchfeasiblefeasiblefeasiblefeasible7MetaNaver
Image Geolocalisation and Visual Place Recognitionfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch7GoogleStanford UniversityPolytechnic University of TurinETH Zurich
Oriented Object Detection in Aerial Imagerystretchstretchfeasiblefeasiblefeasiblefeasiblefeasiblefeasible7Northwestern Polytechnical UniversityWuhan UniversityShanghai Jiao Tong UniversityOpenMMLab
Humanoid Whole-Body Controlfeasiblefeasiblestretchstretchstretchstretchnot feasiblenot feasible6NVIDIAFigure AIGalbotTeslaGoogle DeepMind
Single-Cell and Cell Imaging Foundation Modelsfeasiblefeasiblestretchstretchfeasiblefeasiblenot feasiblenot feasible6Yale UniversityGoogleGoogle DeepMindRecursion
Text-to-Speech for Low-Resource Languagesfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch6MetaAI4Bharat (IIT Madras)ElevenLabsAmpixa
Voice Activity, Turn and Keyword Detectionfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible6NVIDIAPipecat (Daily)Alibaba (Qwen)rednote (Xiaohongshu)SpeechBrain community
Zero-Shot Extraction and Classification with Small Encodersfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible6MetaMicrosoft AI
Diffusion Language Modelsfeasiblefeasiblestretchstretchstretchstretchfeasiblefeasible5Inception (Inception Labs)AppleGoogle
End-to-End Driving Policies with Reasoningfeasiblefeasiblenot feasiblenot feasiblestretchstretchnot feasiblenot feasible5NVIDIATesla
Equivariant Graph Networks for Moleculesfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch5University of AmsterdamTechnische Universitat BerlinTechnical University of MunichMicrosoft ResearchUniversity of Oxford
Interpretability on Small Open Modelsfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch5AnthropicGoogle DeepMindOpen community (Neuronpedia)
Robot Perception for Manipulation and Navigationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible5NVIDIAMetaETH Zurich
Self-Play Reinforcement Learning with Searchfeasiblefeasiblenot feasiblenot feasiblestretchstretchnot feasiblenot feasible5Google DeepMindOpenAI
Speaker Verification Embeddingsfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible5SpeechBrain communityNVIDIAAlibaba (Qwen)
Speech Emotion Recognitionfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible5Alibaba (Qwen)audEERINGSpeechBrain community
Speech-to-Speech Translationfeasiblefeasiblestretchstretchnot feasiblenot feasiblefeasiblefeasible5MetaOpenAINVIDIA
Bioacoustic Classification for Biodiversity Monitoringfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible4GoogleBirdNET team
Imitation Learning from Demonstrationsfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible4Stanford UniversityColumbia UniversityNew York UniversityNVIDIATsinghua University
Kilometre-Scale Generative Downscaling of Weatherfeasiblefeasiblefeasiblefeasiblestretchstretchstretchstretch4NVIDIAMeteoSwiss
Knowledge Graph Reasoning and Relational Foundation Modelsfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible4MilaKumo AIUniversity of AmsterdamIntelPeking University
Model-Based Reinforcement Learning with Learned World Modelsfeasiblefeasiblestretchstretchfeasiblefeasiblestretchstretch4Google DeepMind
Plantation and Natural Forest Mapping from Satellite and Drone Imageryfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible4Google DeepMindGoogleWorld Resources Institute (WRI)EOS Data Analytics (EOSDA)
Speech Recognition for Indonesian and Regional Languagesfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible4Institut Teknologi Bandung (ITB)Universitas IndonesiaJAISTNAISTIndonesian NLP
Table Question Answeringfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible4GoogleMicrosoft AI
Video-to-Audio and Foley Generationfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible4University of Illinois Urbana-ChampaignSonyTencentFunAudioLLMShanghai AI Lab
Deep-Learning Climate Emulationfeasiblefeasiblestretchstretchstretchstretchnot feasiblenot feasible3University of WashingtonNVIDIAGoogle
Hallucination and Grounding Detectionfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible3IBMVectaraBespoke Labs
Mathematical Reasoning and Formal Theorem Provingfeasiblefeasiblestretchstretchstretchstretchnot feasiblenot feasible3DeepSeekTsinghua University
Neural Radiance Fields and Gaussian Splattingfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible3GoogleAppleINRIA
Reinforcement Learning for Algorithm Discoveryfeasiblefeasiblenot feasiblenot feasiblestretchstretchnot feasiblenot feasible3Google DeepMind
Image-Based Virtual Try-onfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible2KAISTOMNIOUS.AISun Yat-sen UniversitySenseTime
Neural Audio Codecs and Audio Tokenisersfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch2GoogleKyutaiNeuphonicHeartMuLa
Parameter-Efficient Fine-Tuning (PEFT)feasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible2Academia and industry broadlyNVIDIA
Differentially Private Language Model Trainingfeasiblefeasiblestretchstretchstretchstretchnot feasiblenot feasible1Google
Evaluation and Benchmark Buildingfeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasiblefeasible1NonprofitsFrontier labsAcademia
Generative Design of Crystalline Materialsfeasiblefeasiblestretchstretchstretchstretchnot feasiblenot feasible1Microsoft AI
Heliophysics Foundation Modelsfeasiblefeasiblestretchstretchfeasiblefeasiblenot feasiblenot feasible1IBMNASA
Latent Dynamics Models of Neural Population Recordingsfeasiblefeasiblestretchstretchstretchstretchnot feasiblenot feasible1GoogleStanford University
Mechanistic Interpretability, Frontier Scalefeasiblefeasiblenot feasiblenot feasiblenot feasiblenot feasiblenot feasiblenot feasible1Anthropic
Alignment and Safety Evaluations (Model Behaviour)feasiblefeasiblefeasiblefeasiblefeasiblefeasiblestretchstretch0Anthropic
Where this comes from
Definitions, anchors and exemplars from T11 of the research notes. Grades: feasible, stretch, not feasible.