Research Papers Kimi K3 is an open Mixture-of-Experts model with 2.8T parameters (104B activated per token), native vision input and a 1-million-token context window, built on Kimi Delta Attention and post-trained with reinforcement learning across general, agentic and coding tasks.
VoiceMem gives duplex speech models a working memory that keeps facts and emotional context across a conversation, so a voice assistant can remember both what you said and how you said it, in real time.
This technical report upgrades LingBot-VLA with a revamped data pipeline of about 60,000 pretraining hours: 50K robot-trajectory hours across 20 embodiments plus 10K egocentric human-video hours. It extends control beyond dual arms to heads, waists, mobile bases and dexterous hands, aiming squarely at the gap between lab robotics and real deployment.
NeoHorse-1 records, for every user turn, the predicted capability demand, the selected service tier and how the interaction went, then converts these records into training examples admitted through structural validation and multi-dimensional semantic checks, feeding them back as the next round of agentic post-training.
RoboDojo is a unified sim-and-real benchmark for generalist robot manipulation policies, with 42 simulation tasks and 18 real-world tasks probing generalization, memory, precision and long-horizon control. It ships parallel Isaac Sim evaluation plus a cloud-accessible real-eval system and a leaderboard the authors populated with 30 policies.
Xiaomi's robotics team presents a vision-language-action foundation model for mobile manipulation, pretrained on more than 100K hours of real-world UMI trajectories with auto-labeled scene-transition language, then post-trained to match embodiments and human imperative prompts. The paper claims clean scaling with more data and parameters, out-of-the-box performance in unseen environments and efficient fine-tuning for dexterous tasks.
Frontis-MA1 ships with OpenMLE, an open full-stack system (task gyms, RL operator learning, evolutionary search) for studying recursive self-improvement. Its 35B meta-evolution agent, post-trained around Draft, Improve, Debug and Crossover program-evolution operators, claims MLE-Bench Lite results the authors say exceed GPT-5.5 plus Codex, with components transferring to held-out NatureBench Lite.
MonkeyOCRv2 pretrains on MonkeyDoc v2, a 113-million-image document corpus spanning 17 languages, combining image-to-text generation with pixel-level stroke reconstruction, so the visual encoder acquires the character-level perception that document tasks demand.
AuK unifies speech generation and editing in a single open-source model: you describe the desired change in natural language, and it either synthesizes new speech or modifies existing audio accordingly.
This reusable skill suite covers the first mile of research ideation: it searches papers across multiple sources, checks whether a novelty claim has already been scooped, and composes end-to-end ideas grounded in evidence from the literature.
When an object leaves the frame, does a video generation model remember it exists? The authors built WROP, 150 cognitive-science tasks across six categories rendered in Blender with randomized lighting, speed, and camera angle — a 1.5M-sample training corpus plus a 300-question exam used to test 14 video models. They then trained PWM-WROP, a 16B world model on their native PyTorch stack for AWS Trainium2, releasing data, exam, scores, and weights. In a blind pairwise Elo study, it ranked first among continuation models and effectively tied the best reference-to-video models.
Spark-to-Paper implements end-to-end paper generation as thirteen composable skills inside an existing coding assistant, with no separate agent platform: it retrieves literature, designs and runs experiments, revises claims to match evidence and produces figures, keeping model judgment separate from mechanically checkable operations.
Realtime-Venus is a proactive full-duplex interaction stack built from two separately trained 9B models: Realtime-Venus-Omni handles audio-visual conversation and Realtime-Venus-Audio handles spoken dialogue, each integrating continuous perception, conversational control and native speech generation on one shared causal timeline. A dual-loop runtime lets the foreground conversation continue while Realtime-Venus-Harness executes delegated tasks asynchronously and folds results back into the live dialogue; the authors report Realtime-Venus-Omni leads evaluated online models on six of eight video benchmarks and Realtime-Venus-Audio beats Gemini 3.1 Live and GPT-4o on all three Full-Duplex-Bench continuation metrics.
LimiX-2 is a tabular foundation model built on Contextual Mechanism Networks: instead of the usual in-context objective of predicting y given x and context tables, it learns the joint structure p(x, y | context) of how the data was generated. It is pretrained with context-conditional masked modeling on synthetic datasets built from structural causal models spanning varied graph structures and mechanisms. The authors report state-of-the-art results against both dataset-specific models and other tabular foundation models on TabArena, TALENT, and BCCO, and show the mechanism-oriented training has a side effect: the model's feature attention encodes direct causal relationships, allowing it to recover causal skeletons from data.
Most open video-language models are either narrowly specialized or only partly reproducible. VideoChat3 goes fully open — code, training strategy, and three curated corpora (Academic2M for general video, LV116K for long-form, OL617K for streaming). Its I3D-ViT encoder inflates a 2D vision transformer into 3D for cheaper spatiotemporal processing, and an adaptive frame-resolution scheme spends compute only where the streaming input needs it. The authors report outperforming earlier open-source models of equal or larger size with only 4B parameters.