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.
Symbolic music models plan composition explicitly but never sound like a finished recording, while audio models produce full songs that keep the composition hidden. YuE2 combines both in a single AR-NAR Mixture-of-Transformers that plans symbolically, expands the plan into semantic music tokens, and renders the full song, using two new representation models (MERT2, SheetSage2) to learn from recordings that lack aligned scores. The authors report a SongBench Global Avg of 6.73 on WildSongBench (6.96 at best-of-8), and in expert listening the best-of-8 samples were preferred over Suno v4.5 and roughly tied with Suno v5; the readable score also makes the output editable by external language-model agents.
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.
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.
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.
Real agents get noisy tool outputs and vague user corrections, and most trained agents just barrel ahead. This paper from Renmin University builds a three-stage pipeline (RSD, ReSFT, ORPO) that first mines reflection behaviors from agent trajectories, distills them into an 8B base model, then sharpens proactive reflection with RL. Built on Qwen3-8B, the authors report gains across six agent suites and stronger robustness on their AgentNoiseBench, which injects realistic noise into tool feedback.
Computer-use models can already read a screen and click, but games—with their visually rich worlds and diverse mechanics—remain hard, and pipelines built on large-scale demonstrations cover only a slice of dynamic game states. Player-1 post-trains such models with synthetic experience instead: it extends the computer-use action space with relative pointer motion, sustained inputs, and composite actions; fine-tunes on 6,044 curated gameplay trajectories paired with task instructions and visual reasoning; and then runs PPO in Player-1-Gym, sixteen synthetic 2D and 3D environments built around control demands shared across games, with programmatically verifiable rewards and randomized observations. The authors report that Player-1-9B and Player-1-35B-A3B beat their own SFT policies on MineDojo-300 by 27.8 and 17.7 percentage points, with Player-1-9B reaching 38.33% success versus 12.63% for the strongest baseline, plus gains on ViZDoom-10 and GameWorld—and the improvements transfer beyond the RL training suite.
OmniCompass evaluates video understanding along three complementary tasks: temporal grounding (when a target appears), spatial grounding (where it is), and question answering (what happens), in two tracks. OmniCompass-VL targets vision-language models with 820 videos and 2,460 instances, while an audio-rich subset forms OmniCompass-AV with 240 videos and 720 instances for models that must combine audio and visual evidence. Every instance is tested under paired text-only and text+image settings, and all data is manually annotated with multi-round verification. Evaluating 50 MLLMs, the authors report open-source models still trail proprietary ones, especially on audio-visual understanding, and that precise temporal and spatial grounding remains hard even for leading models.
Video understanding agents usually need slow chain-of-thought decoding, which is too late for tasks that demand timely choices from a video stream. VideoJev instead casts temporal grounding and multiple-choice video QA as structured decision tasks: it swaps the MLLM's autoregressive answer head for a shared scoring head that rates candidate time spans and text options against the video directly, trained with GRPO. A cached- and joint-inference scheme reuses visual computation across sequential queries on the same video. The authors report mean-mIoU gains of 6.7 and 10.8 points on 4B and 35B-A3B backbones with 2.8-3.1× grounding speedups, and over 100× speedups on video QA while matching Thinking-mode accuracy.
Edge0 is an open-source inference engine that serves a 35B mixture-of-experts model from a single 24GB GPU by loading expert weights from SSD on demand; a trained per-layer prerouter predicts the next layer's expert choices one token ahead so SSD reads overlap with compute, and a recovery LoRA pays back the quality lost to int4 quantization. The authors report 20 tok/s at about 3GiB of peak active memory, within a few points of the fp16 teacher on five public benchmarks.
GroundingPI is a 4B foundation model that outputs points and boxes as quantized coordinates in a shared vocabulary, trained through multimodal and spatial pretraining, supervised fine-tuning, and GRPO reinforcement learning. The authors report state-of-the-art results across 34 grounding benchmarks covering 11 perceptual capabilities, averaging 73.68% against 44 baselines and edging out the larger GPT-6 Astra (71.54%). Used as the visual backbone for downstream tasks, they report it improves robotic manipulation, reaching the best results among evaluated backbones on RoboTwin 2.0 in all four out-of-distribution settings, and lowers open-loop L2 error to 0.296 m on nuScenes for autonomous driving.
Reinforcement learning on diffusion models only rewards the final image, leaving intermediate denoising steps unsupervised. This framework converts the endpoint reward into explicit bounded targets for each sampled query state, training the model on-policy through self-distillation.
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.
Intern-Decision is a family of 0.8B, 2B, and 4B multimodal decision models for settings like games, visual interfaces, and browsers where an agent must pick from a finite set of actions. Its autoregressive masked language modeling objective places one decision placeholder per field in a complete JSON skeleton, so ground-truth symbols appear only in supervision and every field is scored in a single causal pass using the backbone's native LM head. The 4B model reports 90.02% average accuracy across seven evaluation suites, and for a fixed 289-token three-field request on one RTX 4090 it runs at 44.16 ms mean latency. The authors also show temperature scaling reduces calibration error on a small pilot, and routing low-confidence decisions to a larger model lifts accuracy on a hard public benchmark from 73.87% to 87.39%.
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.