Proposes TrOPD, a method that restricts token-level on-policy distillation to regions where teacher supervision is reliable to stabilize training under teacher–student distribution mismatch. Adds outlier handling (clipping, masking, forward-KL) and off-policy guidance; shows consistent gains on math reasoning, code generation and general benchmarks.
Studies small trainable adapters (PEFT) used as persistent personal models on top of large foundation models, analyzing three scaling axes—Scale Up, Scale Down, Scale Out—and introducing MinT, an infrastructure for adapter identity, provenance, evaluation, and serving.
Trains a GPT-style causal Transformer on a 2-billion-frame retargeted motion corpus to enable zero-shot whole-body motion tracking and control. By scaling both data and model capacity, it tracks highly dynamic behaviors while generalizing to unseen motions; accepted to CVPR 2026.
Explores how training recipe — data composition, teacher guidance, and task mixture — shapes few-step distillation for text-to-image generation and instruction-guided image editing; introduces Qwen-Image-Flash and empirical findings that training pipeline organization matters as much as distillation objectives.
AI Dataset2026
Provides ~1M synthetic Salvadoran‑Spanish personas (148k records, ~300M tokens) grounded in 2024 census distributions for demographics, occupations and locations; intended for training/evaluating localized LLMs and synthetic-data workflows. CC BY 4.0, adults only.
Large-scale training corpus for knowledge- and reasoning-intensive video understanding: 315K video reasoning examples over 145K CC-licensed expert-domain videos, with human-in-the-loop chain-of-thought rationales to strengthen post-training for video reasoning. ([arxiv.org](https://arxiv.org/abs/2606.05259))
AI Model2026
Open-weight frontier LLM for agentic reasoning and long-context analysis (up to 1M tokens). Uses a LatentMoE + Mamba-2 hybrid with Multi-Token Prediction and NVFP4 efficiency (550B total / 55B active). Suited for multilingual agents, RAG, and heavy tool-use workloads.
AI Model2026
Multilingual frontier LLM optimized for long-context reasoning and agentic workflows, combining a LatentMoE (Mamba-2 + MoE) hybrid architecture with Multi-Token Prediction and NVFP4 quantization; targeted for NVIDIA GPU deployments and governed by the OpenMDW-1.1 license.
AI Dataset2026
Provides 1,036,431 identity–text–video triplets with per-video JSON annotations and reference keyframes to train and evaluate identity-preserving customized video generation models. Data is drawn from ~320K Pexels HD videos; videos must be downloaded separately per Pexels' terms.
AI Agent Papers2026
Guides LLM-based agents to decompose long-horizon research problems and delegate subtasks to constrained subagents, then fine-tunes models on harness-generated trajectories so delegation decisions become internalized. Reports SearchSwarm-30B-A3B achieving top BrowseComp scores for its scale.
AI Dataset2026
Provides 500+ hours of human whole-body teleoperation demonstrations for humanoid robot learning in real homes, with synchronized video, joint states, action traces and language annotations. Includes 23K+ episodes, fine-grained subtask labels, and raw ROS/MCAP plus compressed LeRobot formats.
AI Dataset2026
Refines large-scale English pretraining corpora by predicting per-instance structured edits (insert, delete, replace) and deterministically applying them to produce cleaner text for LLM training. Provides five ~20B-token refined corpora in parquet with edit metadata and simple loading configs.