Karsten Kuhnke PRO
mindchain
AI & ML interests
Industry Grade Humanoid Synthetic Motion Data Generation, Mechanistic Interpretability Data Generation, Sparse Autoencoders, Edge IOT, Gemma Scope 2, RLHF, Edge AI, Alpa SIM, Alpamayo-R1, Cosmos, Isaac SIM, Isaac LAB, GR00T N1.6, Unreal Engine
Recent Activity
published
a model
about 6 hours ago
mindchain/qwen3-0.6b-finetuned
updated
a model
about 9 hours ago
mindchain/grpo-qwen-test-merged
published
a model
about 9 hours ago
mindchain/grpo-qwen-test-merged
Organizations
NVIDIA GR00T N1 humanoid robotics foundation models for embodied AI, manipulation, and robot learning
NVIDIA Nemotron PII dataset for personally identifiable information detection and privacy-aware NLP
Unitree G1 humanoid robot Dex1 dexterity datasets with mounted camera for manipulation learning
Unitree G1 humanoid robot BrainCo datasets for grasping, object manipulation, and dexterous hand training
-
unitreerobotics/G1_Brainco_GraspOreo_Dataset
Viewer • Updated • 235k • 69 -
unitreerobotics/G1_Brainco_GraspRubiksCube_Dataset
Viewer • Updated • 221k • 92 • 1 -
unitreerobotics/G1_Brainco_PickApple_Dataset
Viewer • Updated • 154k • 38 -
unitreerobotics/G1_Brainco_PickCharger_Dataset
Viewer • Updated • 217k • 33
Hugging Face LeRobot Pi0 legacy version - archived robotics model for reference and compatibility
Open X-Embodiment robotics datasets: cross-platform robot learning for DROID, Kuka, TACO, JACO
Hugging Face LeRobot Pi0 foundation model for robotics: manipulation, navigation, and embodied AI
Hugging Face LeRobot XVLA cross-embodiment vision-language-action models for universal robot control
Higher-order knowledge representations and hypergraph reasoning for agentic AI and scientific discovery
NVIDIA Nemotron Orchestrator 8B for multi-model coordination, task routing, and agentic workflows
Meta RoBERTa pretrained language models for NLP tasks: classification, NER, sentiment analysis
-
FacebookAI/roberta-base
Fill-Mask • 0.1B • Updated • 8.56M • • 561 -
FacebookAI/xlm-roberta-large-finetuned-conll03-german
Token Classification • Updated • 7.17k • • 14 -
FacebookAI/xlm-roberta-large-finetuned-conll02-spanish
Fill-Mask • 0.6B • Updated • 136 • 2 -
FacebookAI/xlm-roberta-large
Fill-Mask • 0.6B • Updated • 5.99M • • 496
NVIDIA Physical AI models for robotics, embodied intelligence, and real-world interaction
Qwen3 Vision-Language reranker models for RAG pipelines and multimodal document retrieval
Meta Research Plan datasets for AI research planning, scientific reasoning, and agent workflows
NVIDIA Clara medical AI models for healthcare, clinical NLP, and medical imaging analysis
NVIDIA Clara molecular models for drug discovery, molecular property prediction, and computational chemistry
NVIDIA Nemotron reranking models for RAG pipelines, search result optimization, and document ranking
NVIDIA Alpamayo-R1 reasoning model for complex problem solving, mathematical reasoning, and chain-of-thought
-
nvidia/Alpamayo-R1-10B
Robotics • Updated • 51.7k • 367 -
nvidia/PhysicalAI-Autonomous-Vehicles
Updated • 380k • 761 -
nvidia/PhysicalAI-Autonomous-Vehicles-NuRec
Updated • 14k • 132 -
Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail
Paper • 2511.00088 • Published • 4
NVIDIA Nemotron speech models for ASR, text-to-speech, and voice AI applications
-
nvidia/nemotron-speech-streaming-en-0.6b
Automatic Speech Recognition • Updated • 9.61k • 475 -
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • Updated • 89.3k • 661 -
nvidia/parakeet_realtime_eou_120m-v1
Updated • 817 • 121 -
nvidia/multitalker-parakeet-streaming-0.6b-v1
Automatic Speech Recognition • Updated • 484 • 80
Open-source GPT models with steering vectors for controllable generation and behavior modification
NVIDIA Cosmos 2 Reason models for world model reasoning, physics simulation, and causal understanding
NVIDIA Cosmos 2.5 Predict models for world simulation, future frame prediction, and physical AI
On-device AI models optimized for smartphone deployment: mobile LLMs, edge inference, and efficient architectures
NVIDIA Nemotron safety datasets for AI alignment, content moderation, and responsible AI training
-
nvidia/Nemotron-AIQ-Agentic-Safety-Dataset-1.0
Viewer • Updated • 10.8k • 1.17k • 14 -
nvidia/Nemotron-Content-Safety-Reasoning-Dataset
Preview • Updated • 157 • 7 -
nvidia/Aegis-AI-Content-Safety-Dataset-2.0
Viewer • Updated • 33.4k • 3.38k • 76 -
nvidia/Nemotron-Content-Safety-Audio-Dataset
Viewer • Updated • 1.93k • 229 • 3
NVIDIA Nemotron vision-language datasets for multimodal training, image understanding, and VLM finetuning
Deep thinking and reasoning models for extended chain-of-thought, deliberative alignment, and complex problem solving
Lightweight code generation models for edge deployment, IDE integration, and fast code completion
YOLO object detection models for real-time computer vision, autonomous systems, and video analytics
Cost-effective coding API providers and affordable LLM endpoints for development and prototyping
Inference Wrapper - Models: Root LLM (The Architect) + Python REPL (The Engine) + Sub LLMs (The Workers) Spaned by querys
NVIDIA Nemotron post-training datasets for RLHF, instruction tuning, and alignment fine-tuning
NVIDIA Nemotron pre-training datasets for large language model training and foundation model development
: Manifold-Constrained Hyper-Connections
Deep research AI agents and models for autonomous literature review, scientific reasoning, and knowledge synthesis
PaddlePaddle PP-StructureV3 for document analysis, table recognition, and intelligent document processing
Circuit sparsity research for neural network pruning, mechanistic interpretability, and efficient model compression
Text-to-motion generation models for human animation, gesture synthesis, and motion capture AI
Text-to-Speech models for voice synthesis, neural TTS, and natural language audio generation
Audio segmentation models based on Meta SAM architecture for sound separation and audio understanding
Mamba/Transformers Combo Hybride
Fully open-source AI models with permissive licenses for commercial use and research
Image-to-3D generation models for single-image 3D reconstruction, mesh generation, and 3D asset creation
Specialized deep research models for domain-specific scientific reasoning and literature analysis
IBM Granite foundation models for enterprise AI, code generation, and multilingual NLP tasks
Lightweight OCR models for edge deployment, mobile text recognition, and efficient document processing
Hierarchical reinforcement learning models for multi-level decision making and complex task decomposition
Top LLMs 2025: ZAI GLM-4.7 (358B) & Moonshot Kimi-K2-Thinking. Next-gen reasoning, code, multilingual. State-of-the-art performance. Production-ready.
Custom JumpReLU Sparse Autoencoders for mechanistic interpretability. T5Gemma-2 SAEs across all layers. AI safety & interpretability research.
NVIDIA Nemotron RL datasets for AI agent training. Web search, workplace tasks, instruction following, structured outputs. RLHF & alignment research.
-
nvidia/Nemotron-RL-knowledge-web_search-mcqa
Viewer • Updated • 2.93k • 118 • 9 -
nvidia/Nemotron-RL-agent-workplace_assistant
Viewer • Updated • 1.8k • 166 • 13 -
nvidia/Nemotron-RL-instruction_following
Preview • Updated • 103 • 11 -
nvidia/Nemotron-RL-instruction_following-structured_outputs
Viewer • Updated • 9.95k • 473 • 34
Custom-trained models by mindchain: reward models & SAEs. Haddock Reward Mini (8B), function-specific SAEs. AI interpretability & alignment research.
JumpReLU SAEs for Gemma 3 interpretability. EleutherAI models, DeepSeek-R1, Pythia SAEs. Mechanistic interpretability & AI safety research.
Function calling: Google FunctionGemma-270m-IT & Mobile Actions dataset (9.65k). Efficient tool use in small LMs. AI agent development.
Google TranslateGemma multilingual translation models supporting 55 languages for neural machine translation and cross-lingual NLP
Unitree Z1 robotic arm datasets for manipulation learning, grasp planning, and arm control training
-
unitreerobotics/Z1_Dual_Dex1_CleanupPencils_Dataset
Viewer • Updated • 133k • 47 • 2 -
unitreerobotics/Z1_Dual_Dex1_FoldClothes_Dataset
Viewer • Updated • 293k • 59 • 2 -
unitreerobotics/Z1_Dual_Dex1_PourCoffee_Dataset
Viewer • Updated • 443k • 37 • 1 -
unitreerobotics/Z1_Dual_Dex1_StackBox_Dataset
Viewer • Updated • 117k • 105 • 4
Unitree G1 humanoid robot Dex3 advanced dexterity datasets for fine-grained manipulation tasks
-
unitreerobotics/G1_Dex3_BlockStacking_Dataset
Viewer • Updated • 281k • 170 • 2 -
unitreerobotics/G1_Dex3_CameraPackaging_Dataset
Viewer • Updated • 256k • 50 -
unitreerobotics/G1_Dex3_GraspSquare_Dataset
Viewer • Updated • 281k • 85 -
unitreerobotics/G1_Dex3_ObjectPlacement_Dataset
Viewer • Updated • 98.3k • 71 • 4
Unitree UnifoLM World Model Agent for robot learning, action prediction, and embodied AI planning
Hugging Face LeRobot Pi0.5 intermediate robotics model with improved action generation capabilities
Hugging Face LeRobot SmolVLA compact vision-language-action model for efficient robot control
Hugging Face LeRobot Behavior-1K large-scale robotics benchmark for diverse manipulation tasks
Atlas Reinforcement Learning - How to put the Intelligence into the Architecture?!
Optimized LLMs for dual NVIDIA RTX 6000 GPU setup - 96GB VRAM configurations for local inference
Hugging Face LeRobot Pi0Fast optimized robotics models for real-time inference and fast action generation
Google Embedding Gemma models for semantic search, RAG applications, and text embeddings
NVIDIA Thor SoC with Raspberry Pi and OAK-4D stereo camera for edge robotics and embodied AI
Qwen3 Vision-Language embedding models for multimodal RAG, semantic search, and vector databases
NVIDIA Llama Nemotron content safety models for toxicity detection and safe AI deployment
NVIDIA Clara biology models for genomics, protein structure, and computational biology research
NVIDIA Clara medical AI for clinical NLP, radiology analysis, and healthcare decision support
NVIDIA Nemotron embedding models for RAG, semantic search, and vector database applications
Diffusion Transformer models for multimodal video and audio generation, synthesis, and editing
-
Lightricks/LTX-2
Image-to-Video • Updated • 1.97M • • 1.56k -
Masked Audio Generation using a Single Non-Autoregressive Transformer
Paper • 2401.04577 • Published • 45 -
LTX-2: Efficient Joint Audio-Visual Foundation Model
Paper • 2601.03233 • Published • 154 -
YOLO-World: Real-Time Open-Vocabulary Object Detection
Paper • 2401.17270 • Published • 43
NVIDIA Nemotron Cascade for multi-stage inference, model routing, and efficient LLM deployment
Google Gemma 3 LiteRT models optimized for TensorFlow Lite runtime and mobile edge deployment
NVIDIA Cosmos 2.5 Transfer models for domain adaptation, style transfer, and video generation
Robotics foundation models, datasets, and research for embodied AI, manipulation, and autonomous systems
NVIDIA Nemotron reinforcement learning datasets from NeMo Gym for agent training and RLHF
-
nvidia/Nemotron-RL-knowledge-web_search-mcqa
Viewer • Updated • 2.93k • 118 • 9 -
nvidia/Nemotron-RL-agent-workplace_assistant
Viewer • Updated • 1.8k • 166 • 13 -
nvidia/Nemotron-RL-instruction_following
Preview • Updated • 103 • 11 -
nvidia/Nemotron-RL-instruction_following-structured_outputs
Viewer • Updated • 9.95k • 473 • 34
NVIDIA Nemotron RAG datasets for retrieval-augmented generation, document QA, and knowledge grounding
Google Gemma 3N mobile multimodal models for on-device vision-language tasks and efficient edge deployment
Compact reasoning models for efficient chain-of-thought inference on resource-constrained devices
Self-Correcting Delta Transformer - DDL provides the Hardware mechanism (The Erazor), NL solves the software problem.
Meta SAM Segment Anything models for zero-shot image segmentation, object detection, and visual understanding
Ultra-compact LLMs for edge deployment: sub-1B parameter models with strong performance for IoT and mobile
NVIDIA Nemotron persona datasets for character AI, personality modeling, and conversational agent training
LLM as a judge
-
nvidia/Llama-3.3-Nemotron-70B-Reward-Principle
Text Generation • 71B • Updated • 217 • 6 -
nvidia/Qwen3-Nemotron-32B-GenRM-Principle
Text Generation • 33B • Updated • 313 • 13 -
nvidia/Qwen3-Nemotron-32B-RLBFF
Text Generation • 33B • Updated • 21 • 27 -
nvidia/Qwen3-Nemotron-8B-BRRM
Text Generation • Updated • 35 • 8
Text and multimodal embedding models for semantic search, RAG pipelines, and vector similarity applications
Edge-optimized translation models for on-device multilingual NLP and low-latency language translation
Qwen long-context reasoning models for extended chain-of-thought, complex problem solving, and mathematical reasoning
Optical Character Recognition models for text extraction, document digitization, and scene text detection
Iquenst LoopCoder models for iterative code generation, self-refinement, and automated debugging
Automatic Speech Recognition models for transcription, voice AI, and multilingual speech-to-text
Mobile app engine models for on-device AI, app development automation, and mobile-first ML
Hybrid attention models combining local and global attention for efficient long-context processing
Mamba/Transformers Combo Hybride
Byte-level language models for tokenizer-free NLP, multilingual text, and raw byte processing
Video analysis models for action recognition, temporal understanding, and video content classification
Diffusion-based language models for text generation, discrete diffusion, and non-autoregressive NLP
Video generation models for text-to-video, image-to-video, and AI video synthesis
Graphics and visual computing models for rendering, image synthesis, and computer graphics AI
Meta VL-JEPA Vision-Language Joint Embedding Predictive Architecture for video understanding
Google Gemma Scope 2: JumpReLU SAEs for Gemma 3 interpretability. 270M PT/IT, 1B PT variants. Neuronpedia integration. Mechanistic analysis.
Quantized Gemma 3 models: QAT for efficient deployment. Gemma-3-27B-IT Q4. Low memory, fast inference. Edge & production-ready LLMs.
NVIDIA Nemotron reward models: 340B, 8B BRRM, 70B/32B principle-based. RLHF training, preference learning, AI alignment research.
T5Gemma-2 encoder-decoder models: 270M, 1B, 4B sizes. Text-to-text, summarization, translation. Google's architecture for structured generation.
-
google/t5gemma-2-270m-270m
Image-Text-to-Text • 0.8B • Updated • 17.9k • 183 -
google/t5gemma-2-4b-4b
Image-Text-to-Text • 9B • Updated • 13.9k • 141 -
google/t5gemma-2-1b-1b
Image-Text-to-Text • 2B • Updated • 20.9k • 69 -
T5Gemma 2: Seeing, Reading, and Understanding Longer
Paper • 2512.14856 • Published • 3
Hybrid Mamba + Transformer architectures. NVIDIA Nemotron-3-Nano-30B-A3B (32B). BF16 & GGUF. Efficient long-context & in-context learning.
-
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
Text Generation • 32B • Updated • 1.03M • 640 -
bartowski/nvidia_Nemotron-3-Nano-30B-A3B-GGUF
Text Generation • 32B • Updated • 1.78k • 9 -
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-Base-BF16
Text Generation • 32B • Updated • 71.1k • 107 -
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8
Text Generation • 32B • Updated • 1.08M • • 284
NVIDIA GR00T N1 humanoid robotics foundation models for embodied AI, manipulation, and robot learning
Google TranslateGemma multilingual translation models supporting 55 languages for neural machine translation and cross-lingual NLP
NVIDIA Nemotron PII dataset for personally identifiable information detection and privacy-aware NLP
Unitree Z1 robotic arm datasets for manipulation learning, grasp planning, and arm control training
-
unitreerobotics/Z1_Dual_Dex1_CleanupPencils_Dataset
Viewer • Updated • 133k • 47 • 2 -
unitreerobotics/Z1_Dual_Dex1_FoldClothes_Dataset
Viewer • Updated • 293k • 59 • 2 -
unitreerobotics/Z1_Dual_Dex1_PourCoffee_Dataset
Viewer • Updated • 443k • 37 • 1 -
unitreerobotics/Z1_Dual_Dex1_StackBox_Dataset
Viewer • Updated • 117k • 105 • 4
Unitree G1 humanoid robot Dex1 dexterity datasets with mounted camera for manipulation learning
Unitree G1 humanoid robot Dex3 advanced dexterity datasets for fine-grained manipulation tasks
-
unitreerobotics/G1_Dex3_BlockStacking_Dataset
Viewer • Updated • 281k • 170 • 2 -
unitreerobotics/G1_Dex3_CameraPackaging_Dataset
Viewer • Updated • 256k • 50 -
unitreerobotics/G1_Dex3_GraspSquare_Dataset
Viewer • Updated • 281k • 85 -
unitreerobotics/G1_Dex3_ObjectPlacement_Dataset
Viewer • Updated • 98.3k • 71 • 4
Unitree G1 humanoid robot BrainCo datasets for grasping, object manipulation, and dexterous hand training
-
unitreerobotics/G1_Brainco_GraspOreo_Dataset
Viewer • Updated • 235k • 69 -
unitreerobotics/G1_Brainco_GraspRubiksCube_Dataset
Viewer • Updated • 221k • 92 • 1 -
unitreerobotics/G1_Brainco_PickApple_Dataset
Viewer • Updated • 154k • 38 -
unitreerobotics/G1_Brainco_PickCharger_Dataset
Viewer • Updated • 217k • 33
Unitree UnifoLM World Model Agent for robot learning, action prediction, and embodied AI planning
Hugging Face LeRobot Pi0 legacy version - archived robotics model for reference and compatibility
Hugging Face LeRobot Pi0.5 intermediate robotics model with improved action generation capabilities
Open X-Embodiment robotics datasets: cross-platform robot learning for DROID, Kuka, TACO, JACO
Hugging Face LeRobot SmolVLA compact vision-language-action model for efficient robot control
Hugging Face LeRobot Pi0 foundation model for robotics: manipulation, navigation, and embodied AI
Hugging Face LeRobot Behavior-1K large-scale robotics benchmark for diverse manipulation tasks
Hugging Face LeRobot XVLA cross-embodiment vision-language-action models for universal robot control
Atlas Reinforcement Learning - How to put the Intelligence into the Architecture?!
Higher-order knowledge representations and hypergraph reasoning for agentic AI and scientific discovery
Optimized LLMs for dual NVIDIA RTX 6000 GPU setup - 96GB VRAM configurations for local inference
NVIDIA Nemotron Orchestrator 8B for multi-model coordination, task routing, and agentic workflows
Hugging Face LeRobot Pi0Fast optimized robotics models for real-time inference and fast action generation
Meta RoBERTa pretrained language models for NLP tasks: classification, NER, sentiment analysis
-
FacebookAI/roberta-base
Fill-Mask • 0.1B • Updated • 8.56M • • 561 -
FacebookAI/xlm-roberta-large-finetuned-conll03-german
Token Classification • Updated • 7.17k • • 14 -
FacebookAI/xlm-roberta-large-finetuned-conll02-spanish
Fill-Mask • 0.6B • Updated • 136 • 2 -
FacebookAI/xlm-roberta-large
Fill-Mask • 0.6B • Updated • 5.99M • • 496
Google Embedding Gemma models for semantic search, RAG applications, and text embeddings
NVIDIA Physical AI models for robotics, embodied intelligence, and real-world interaction
NVIDIA Thor SoC with Raspberry Pi and OAK-4D stereo camera for edge robotics and embodied AI
Qwen3 Vision-Language reranker models for RAG pipelines and multimodal document retrieval
Qwen3 Vision-Language embedding models for multimodal RAG, semantic search, and vector databases
Meta Research Plan datasets for AI research planning, scientific reasoning, and agent workflows
NVIDIA Llama Nemotron content safety models for toxicity detection and safe AI deployment
NVIDIA Clara medical AI models for healthcare, clinical NLP, and medical imaging analysis
NVIDIA Clara biology models for genomics, protein structure, and computational biology research
NVIDIA Clara molecular models for drug discovery, molecular property prediction, and computational chemistry
NVIDIA Clara medical AI for clinical NLP, radiology analysis, and healthcare decision support
NVIDIA Nemotron reranking models for RAG pipelines, search result optimization, and document ranking
NVIDIA Nemotron embedding models for RAG, semantic search, and vector database applications
NVIDIA Alpamayo-R1 reasoning model for complex problem solving, mathematical reasoning, and chain-of-thought
-
nvidia/Alpamayo-R1-10B
Robotics • Updated • 51.7k • 367 -
nvidia/PhysicalAI-Autonomous-Vehicles
Updated • 380k • 761 -
nvidia/PhysicalAI-Autonomous-Vehicles-NuRec
Updated • 14k • 132 -
Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail
Paper • 2511.00088 • Published • 4
Diffusion Transformer models for multimodal video and audio generation, synthesis, and editing
-
Lightricks/LTX-2
Image-to-Video • Updated • 1.97M • • 1.56k -
Masked Audio Generation using a Single Non-Autoregressive Transformer
Paper • 2401.04577 • Published • 45 -
LTX-2: Efficient Joint Audio-Visual Foundation Model
Paper • 2601.03233 • Published • 154 -
YOLO-World: Real-Time Open-Vocabulary Object Detection
Paper • 2401.17270 • Published • 43
NVIDIA Nemotron speech models for ASR, text-to-speech, and voice AI applications
-
nvidia/nemotron-speech-streaming-en-0.6b
Automatic Speech Recognition • Updated • 9.61k • 475 -
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • Updated • 89.3k • 661 -
nvidia/parakeet_realtime_eou_120m-v1
Updated • 817 • 121 -
nvidia/multitalker-parakeet-streaming-0.6b-v1
Automatic Speech Recognition • Updated • 484 • 80
NVIDIA Nemotron Cascade for multi-stage inference, model routing, and efficient LLM deployment
Open-source GPT models with steering vectors for controllable generation and behavior modification
Google Gemma 3 LiteRT models optimized for TensorFlow Lite runtime and mobile edge deployment
NVIDIA Cosmos 2 Reason models for world model reasoning, physics simulation, and causal understanding
NVIDIA Cosmos 2.5 Transfer models for domain adaptation, style transfer, and video generation
NVIDIA Cosmos 2.5 Predict models for world simulation, future frame prediction, and physical AI
Robotics foundation models, datasets, and research for embodied AI, manipulation, and autonomous systems
On-device AI models optimized for smartphone deployment: mobile LLMs, edge inference, and efficient architectures
NVIDIA Nemotron reinforcement learning datasets from NeMo Gym for agent training and RLHF
-
nvidia/Nemotron-RL-knowledge-web_search-mcqa
Viewer • Updated • 2.93k • 118 • 9 -
nvidia/Nemotron-RL-agent-workplace_assistant
Viewer • Updated • 1.8k • 166 • 13 -
nvidia/Nemotron-RL-instruction_following
Preview • Updated • 103 • 11 -
nvidia/Nemotron-RL-instruction_following-structured_outputs
Viewer • Updated • 9.95k • 473 • 34
NVIDIA Nemotron safety datasets for AI alignment, content moderation, and responsible AI training
-
nvidia/Nemotron-AIQ-Agentic-Safety-Dataset-1.0
Viewer • Updated • 10.8k • 1.17k • 14 -
nvidia/Nemotron-Content-Safety-Reasoning-Dataset
Preview • Updated • 157 • 7 -
nvidia/Aegis-AI-Content-Safety-Dataset-2.0
Viewer • Updated • 33.4k • 3.38k • 76 -
nvidia/Nemotron-Content-Safety-Audio-Dataset
Viewer • Updated • 1.93k • 229 • 3
NVIDIA Nemotron RAG datasets for retrieval-augmented generation, document QA, and knowledge grounding
NVIDIA Nemotron vision-language datasets for multimodal training, image understanding, and VLM finetuning
Google Gemma 3N mobile multimodal models for on-device vision-language tasks and efficient edge deployment
Deep thinking and reasoning models for extended chain-of-thought, deliberative alignment, and complex problem solving
Compact reasoning models for efficient chain-of-thought inference on resource-constrained devices
Lightweight code generation models for edge deployment, IDE integration, and fast code completion
Self-Correcting Delta Transformer - DDL provides the Hardware mechanism (The Erazor), NL solves the software problem.
YOLO object detection models for real-time computer vision, autonomous systems, and video analytics
Meta SAM Segment Anything models for zero-shot image segmentation, object detection, and visual understanding
Cost-effective coding API providers and affordable LLM endpoints for development and prototyping
Ultra-compact LLMs for edge deployment: sub-1B parameter models with strong performance for IoT and mobile
Inference Wrapper - Models: Root LLM (The Architect) + Python REPL (The Engine) + Sub LLMs (The Workers) Spaned by querys
NVIDIA Nemotron persona datasets for character AI, personality modeling, and conversational agent training
NVIDIA Nemotron post-training datasets for RLHF, instruction tuning, and alignment fine-tuning
LLM as a judge
-
nvidia/Llama-3.3-Nemotron-70B-Reward-Principle
Text Generation • 71B • Updated • 217 • 6 -
nvidia/Qwen3-Nemotron-32B-GenRM-Principle
Text Generation • 33B • Updated • 313 • 13 -
nvidia/Qwen3-Nemotron-32B-RLBFF
Text Generation • 33B • Updated • 21 • 27 -
nvidia/Qwen3-Nemotron-8B-BRRM
Text Generation • Updated • 35 • 8
NVIDIA Nemotron pre-training datasets for large language model training and foundation model development
Text and multimodal embedding models for semantic search, RAG pipelines, and vector similarity applications
: Manifold-Constrained Hyper-Connections
Edge-optimized translation models for on-device multilingual NLP and low-latency language translation
Deep research AI agents and models for autonomous literature review, scientific reasoning, and knowledge synthesis
Qwen long-context reasoning models for extended chain-of-thought, complex problem solving, and mathematical reasoning
PaddlePaddle PP-StructureV3 for document analysis, table recognition, and intelligent document processing
Optical Character Recognition models for text extraction, document digitization, and scene text detection
Circuit sparsity research for neural network pruning, mechanistic interpretability, and efficient model compression
Iquenst LoopCoder models for iterative code generation, self-refinement, and automated debugging
Text-to-motion generation models for human animation, gesture synthesis, and motion capture AI
Automatic Speech Recognition models for transcription, voice AI, and multilingual speech-to-text
Text-to-Speech models for voice synthesis, neural TTS, and natural language audio generation
Mobile app engine models for on-device AI, app development automation, and mobile-first ML
Audio segmentation models based on Meta SAM architecture for sound separation and audio understanding
Hybrid attention models combining local and global attention for efficient long-context processing
Mamba/Transformers Combo Hybride
Mamba/Transformers Combo Hybride
Fully open-source AI models with permissive licenses for commercial use and research
Byte-level language models for tokenizer-free NLP, multilingual text, and raw byte processing
Image-to-3D generation models for single-image 3D reconstruction, mesh generation, and 3D asset creation
Video analysis models for action recognition, temporal understanding, and video content classification
Specialized deep research models for domain-specific scientific reasoning and literature analysis
Diffusion-based language models for text generation, discrete diffusion, and non-autoregressive NLP
IBM Granite foundation models for enterprise AI, code generation, and multilingual NLP tasks
Video generation models for text-to-video, image-to-video, and AI video synthesis
Lightweight OCR models for edge deployment, mobile text recognition, and efficient document processing
Graphics and visual computing models for rendering, image synthesis, and computer graphics AI
Hierarchical reinforcement learning models for multi-level decision making and complex task decomposition
Meta VL-JEPA Vision-Language Joint Embedding Predictive Architecture for video understanding
Top LLMs 2025: ZAI GLM-4.7 (358B) & Moonshot Kimi-K2-Thinking. Next-gen reasoning, code, multilingual. State-of-the-art performance. Production-ready.
Google Gemma Scope 2: JumpReLU SAEs for Gemma 3 interpretability. 270M PT/IT, 1B PT variants. Neuronpedia integration. Mechanistic analysis.
Custom JumpReLU Sparse Autoencoders for mechanistic interpretability. T5Gemma-2 SAEs across all layers. AI safety & interpretability research.
Quantized Gemma 3 models: QAT for efficient deployment. Gemma-3-27B-IT Q4. Low memory, fast inference. Edge & production-ready LLMs.
NVIDIA Nemotron RL datasets for AI agent training. Web search, workplace tasks, instruction following, structured outputs. RLHF & alignment research.
-
nvidia/Nemotron-RL-knowledge-web_search-mcqa
Viewer • Updated • 2.93k • 118 • 9 -
nvidia/Nemotron-RL-agent-workplace_assistant
Viewer • Updated • 1.8k • 166 • 13 -
nvidia/Nemotron-RL-instruction_following
Preview • Updated • 103 • 11 -
nvidia/Nemotron-RL-instruction_following-structured_outputs
Viewer • Updated • 9.95k • 473 • 34
NVIDIA Nemotron reward models: 340B, 8B BRRM, 70B/32B principle-based. RLHF training, preference learning, AI alignment research.
Custom-trained models by mindchain: reward models & SAEs. Haddock Reward Mini (8B), function-specific SAEs. AI interpretability & alignment research.
T5Gemma-2 encoder-decoder models: 270M, 1B, 4B sizes. Text-to-text, summarization, translation. Google's architecture for structured generation.
-
google/t5gemma-2-270m-270m
Image-Text-to-Text • 0.8B • Updated • 17.9k • 183 -
google/t5gemma-2-4b-4b
Image-Text-to-Text • 9B • Updated • 13.9k • 141 -
google/t5gemma-2-1b-1b
Image-Text-to-Text • 2B • Updated • 20.9k • 69 -
T5Gemma 2: Seeing, Reading, and Understanding Longer
Paper • 2512.14856 • Published • 3
JumpReLU SAEs for Gemma 3 interpretability. EleutherAI models, DeepSeek-R1, Pythia SAEs. Mechanistic interpretability & AI safety research.
Hybrid Mamba + Transformer architectures. NVIDIA Nemotron-3-Nano-30B-A3B (32B). BF16 & GGUF. Efficient long-context & in-context learning.
-
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
Text Generation • 32B • Updated • 1.03M • 640 -
bartowski/nvidia_Nemotron-3-Nano-30B-A3B-GGUF
Text Generation • 32B • Updated • 1.78k • 9 -
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-Base-BF16
Text Generation • 32B • Updated • 71.1k • 107 -
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8
Text Generation • 32B • Updated • 1.08M • • 284
Function calling: Google FunctionGemma-270m-IT & Mobile Actions dataset (9.65k). Efficient tool use in small LMs. AI agent development.