New releases apply LLMs and AI agents to genomic prediction, scientific discovery, supply-chain optimization, kernel engineering, robotics and autonomous driving. The common thread is moving beyond chatbots to real-world systems and physical workloads.
A multi-agent framework generates QUBO formulations from natural-language descriptions and uses test cases to correct them. On QUBOBench, 100 problems across 12 domains, it reaches 68% accuracy, beating a single-call baseline by 22%. The result shift...
NVIDIA has built a command centre for supply chain allocation based on Palantir Foundry and cuOpt. The system combines mixed-integer linear programming with a Nemotron 3.5 Lightning LLM with 30 billion total parameters, three billion active, fine-tun...
M3-Former combines static attributes and navigation intent encoded by an LLM with dynamic features, using a dual-granularity MoE and a steering-specific loss. On a Danish AIS dataset it reduces 4-hour ADE and FDE by 4.4% and 5.1% against the stronges...
A public dataset records 558 complete AI scientific agent trajectories with thoughts, tool calls, errors, and confidence. Traditional benchmarks look only at final outputs; here comparable frontier models reveal different error profiles. Claude Opus ...
An ARM engineer used LLMs to find bottlenecks in Linux kernel compilation. The model generated a lot of code, much of it low quality, but identified room for parallelization that enables faster builds. The value is not in the generated code but in th...
Qwen has published an open-weight model derived from a 4B checkpoint and fine-tuned for driving. The full BF16 checkpoint is 9B. It signals Chinese labs pushing toward autonomous driving with open models, with implications for local inference, data s...
DeepMind has released AlphaGenome Atlas, a public repository of precomputed predictions for nine billion possible single-letter changes to the human genome. It removes the need to write code or run the model, adds an impact score, and is available fo...
Arm brings together more than 80 partners to define common standards for robotics and physical AI. The Robotics Capability Framework introduces capability tiers similar to SAE levels, tying use cases to latency, compute placement, memory, power, and ...
