📄 Trending AI Research

Vidu S2 introduces real-time interactive avatar and video editing models that support high-resolution spatial video generation and dynamic reference updates.

Waypoint-OPD improves policy-shift distillation by gating endpoint shifts using directional agreement between early and late teacher policy changes, boosting reasoning performance with minimal overhead.

A 150M-parameter reasoning model using recurrent latent reasoning and in-context learning achieves a new cost-accuracy frontier on ARC-AGI-1.

DisCo is a research agent that distills operational knowledge into reusable skills, significantly improving autonomous ML research performance across benchmarks.

StudentSim trains personalized student simulators from sparse data to mirror learner responses and adapt to tutor guidance, outperforming existing models across chess, writing, and math.

World Model RL replaces costly environment execution with a learned world model and applies debiasing and denoising to accelerate post-training of autonomous research agents.

Atria Dawn Preview is a foundation agentic language model trained through verified tool interactions that achieves strong benchmark results and demonstrates a shift toward human-AI project-level collaboration in scientific research.

NeoHorse-1 uses agentic post-training with intelligent routing, structured feedback loops, and curriculum-based distillation to improve model capabilities across agent benchmarks.

Compile by training converts natural-language specifications into reusable neural functions by distilling teacher-generated examples into small adapters, enabling efficient deployment without remote model dependencies.

StateM is a runtime system that improves long-horizon agent execution through durable states, recoverable runbooks, and enforceable procedural controls without altering model weights.

Qwen-Drive-1.0 is a vision-language foundation model for autonomous driving that unifies 3D perception, visual question answering, and motion planning via shared representations and staged training.

MJev-Spatial is a fast multimodal decision model that unifies spatial reasoning tasks into a single-pass classification framework, improving speed and accuracy over autoregressive vision-language models.

Kimi K3 is a 2.8-trillion-parameter open-source AI model that uses a Mixture-of-Experts design to activate only 104 billion parameters per token, making it efficient despite its size. It introduces new attention mechanisms and a stable expert-routing method to improve performance across long sequences and deep networks, achieving roughly 2.5x better scaling efficiency than its predecessor. The model excels at long-horizon coding, agentic tasks, reasoning, and vision, outperforming all other open models and most proprietary ones, though it still trails the top closed models like Claude Fable 5 and GPT-5.6 Sol. Its release as open weights aims to accelerate research and deployment of frontier-level AI.

OpenGCUA introduces an open generalist computer-use agent that integrates visual and programmatic interfaces, harness-to-model decision transfer, and code execution to improve task performance and efficiency across models.

NCP-ArchPreview is a large latent-space language model that jointly trains next-token and next-concept prediction to improve pretraining efficiency and downstream performance.

Macaron-V1 is an open agent-model family that uses a Mixture-of-LoRA architecture and recursive self-improvement to enable continual learning and collaboration across specialized tasks.

ZGCM-1 is a 7B open foundation model that combines internal reasoning with external tool use, trained via efficient architecture-system co-design, progressive long-context scaling, and autonomous agent workflows to achieve strong reasoning and efficiency.

Composing complementary continual learning mechanisms substantially improves long-horizon memorization in language models by combining data, function, and weight anchors with merged low-rank updates.

This paper introduces Qwen-UI-Agent, a foundation model designed to act as a general-purpose executor across mobile, desktop, web, and search interfaces. Unlike typical GUI agents that operate in isolated sandboxes, it runs on real devices and unifies GUI clicks with command-line actions, generating multiple steps per turn to handle long, complex workflows. A key innovation is an automated data flywheel where the agent itself builds tasks, diagnoses its own failures, and plans improvements, plus online reinforcement learning on trajectories over 100 steps using 10,000+ parallel environments. The headline results show state-of-the-art mobile performance (82.1% on MobileWorld, 92.2% on MobileWorld-Real) and competitive computer/browser scores (79.5% on OSWorld-Verified, 73.6% on WebArena) against frontier models like GPT-5.6 and Gemini 3.1 Pro. The work matters because it pushes agents from scripted demos toward reliable, self-improving operation on actual devices, a prerequisite for practical digital assistants.

AskChem changes chemistry literature search from returning ranked papers to returning atomic, provenance-carrying claims—each tied to a source DOI and a verbatim quote or evidence locator. It indexes 2.4M claims from 147K papers and offers multiple access paths: a faceted taxonomy for browsing, an evidence graph linking claims, and a living taxonomy organized by scientific principles, plus REST, SDK, and MCP interfaces for AI agents. In their benchmark, grounding a GPT-5.5 reader in AskChem achieved 100% resolvable DOIs versus 88.3% without retrieval, and the highest citation density among five systems tested. This matters because it directly addresses the manual assembly and verification burden in cross-paper synthesis, for both humans and AI agents.

FDE-Gym evaluates AI agents on realistic customer delivery tasks requiring clarification of ambiguous requirements and discovery of unstated constraints, revealing a substantial performance gap compared to well-specified benchmarks.

Terminal-Universe reconstructs executable workspaces from agent trajectories to synthesize diverse training tasks and improves post-training performance through supervised fine-tuning.

Dream-RSI enables scalable recursive self-improvement by using historical discovery replay to evaluate exploration policies offline, reducing costly online evaluations.

This paper introduces the concept of a 'memory foundation model,' arguing that current AI agents rely on external memory systems (like vector databases) rather than building memory directly into the model's weights and computation. The authors propose Metis, a prototype that maintains a persistent, evolving memory state inside the model's backbone, updated via a simple forward pass without gradient updates. This design allows historical information to be compressed into the model and accessed through a special 'memory attention' mechanism, keeping all learned weights frozen during inference. The key result is that Metis demonstrates native memory capabilities—storing and retrieving information autonomously—while offering architectural and efficiency advantages over external memory approaches, though the paper also analyzes its current limitations.

A new benchmark for AI-generated video detection reveals that current detectors fail to generalize across realistic crisis-related videos and become less reliable as content spreads socially.

EnvHarness and EnvRigger dynamically reshape static environments via programmable plugins to target agent weaknesses and improve reinforcement learning co-evolution.

OpenART introduces a scalable red-teaming arena with evolving stateful environments to evaluate long-horizon AI agent safety, using the EMHA attack policy to expose increasing failure rates as task complexity grows.

SenseNova-U1.5 is an 8B native unified multimodal model that performs visual understanding, reasoning, and generation without encoders or VAEs, achieving high fidelity and instruction following through patch reconstruction, curated data, expert optimization, and on-policy distillation.

Spark-to-Paper is a lightweight, composable workflow inside coding assistants that generates research papers by separating planning from reporting, enforcing evidence-based claim revision, and using integrity checks to reduce fabrication.

HarnessDev evaluates agents by measuring their ability to build and iteratively improve execution infrastructure rather than final task outputs, revealing that self-built harnesses vary widely in capability and efficiency and transfer poorly across models.

Parameter-efficient fine-tuning with low-rank adapters enables efficient adaptation of very large multimodal backbones for brain encoding, improving performance and interpretability while reducing computational cost.

LLaDA-Image unifies a 6B diffusion transformer with a frozen vision-language module, using image-only pre-training and a Muon optimizer to generate photorealistic images with precise editing, and is distilled into a fast 2-4 step variant that achieves state-of-the-art open-source results.

This paper introduces a method to automatically generate large volumes of training data for AI agents that perform long, multi-step tasks in terminal environments (like running commands in a shell). The key problem is that such data is normally hand-crafted and costs hundreds to thousands of dollars per task because the instruction, environment, and solution must all be consistent. The new approach, called Recursive Synthetic Terminal Tasks (RST), starts with a few verified tasks and repeatedly extends their solutions, then automatically updates the instructions and checks to match, validating each new task in a sandbox. Over 15 rounds, it produced 37,484 tasks at about $0.05 each, with difficulty rising sharply—median solution length grew from 67 to 374 lines, and the best model's success rate dropped from 90% to 2.5%. Training models on these synthetic tasks improved their performance on standard terminal-agent benchmarks by up to 41% relative, and the process shows no signs of plateauing, suggesting it could scale further.

GPT-6-Astra achieves high performance across diverse robotic tasks via in-context learning, yet real-world deployment reveals gaps in demonstration cost, execution time, and practical embodied intelligence.