Mila

Context compaction in agentic RL can run up to 5x faster simply by streaming the KV cache instead of flushing it—accidentally turning standard LLMs into recurrent agents that preserve evicted context purely through RL.

Reacting to sudden physical hazards is both a meaningful test of embodied intelligence and a hard requirement for deploying multimodal large language models (MLLMs) as the decision core of household robots, and ReactHuman is the first physics-grounded benchmark for human-like reactive decision-making.

A single policy model outperforms complex tree search methods in chemistry tool learning, achieving notable improvements in performance metrics.

Evolving agent harnesses can boost performance by up to 35% in enterprise environments, even with fixed model weights.

Adaptive temporal discounting can dramatically enhance reinforcement learning's efficiency and flexibility, outperforming traditional fixed discount methods.

SOGR1 achieves state-of-the-art performance in knowledge graph question answering while being significantly more efficient than larger models.

Even advanced LLMs struggle to prevent privacy breaches in multi-user settings, exposing critical data spillage risks that current benchmarks overlook.

Reusable fixing transformations can achieve a 94.3% compilable transformation rate, revolutionizing how we handle breaking API changes across multiple projects.

Conditioning LLMs on human privacy judgments leads to a remarkable increase in alignment with user expectations, showcasing a new standard for agent training.

N-version programming with coding agents not only mirrors historical failures but also shows a dramatic reduction in errors through diversity, challenging assumptions about AI reliability.

AuditFlow achieves over 82% accuracy in structured financial audits by leveraging a unique symbolic environment that outperforms traditional methods by nearly 15 points.

MobEvolve outperforms traditional methods by evolving its logic through targeted updates, achieving unprecedented fidelity and interpretability in human mobility generation.

How you represent a plan matters more than which LLM you use when building robust web agents.

Even state-of-the-art vision-language models frequently lie and hallucinate when playing social deduction games, raising serious questions about their reliability in real-world applications requiring grounded reasoning.

As AI agents scale and interact, the Foundation Protocol offers a coordination layer that prioritizes accountability and governance, ensuring that the future of human-AI collaboration remains open and governable.

Forget fixed teams: this new reinforcement learning framework lets agents spawn new teammates on the fly, unlocking dynamic strategies previously impossible.