vLLM v0.31.0 ships as Microsoft remakes Windows for agents

The hallucination gate reshuffles the leaderboard: Muse Spark 1.3 would lead at 26.7% all-pass without it, but two thirds of its passes contain material hallucinations, while GPT-6 Astra keeps nearly all its passes and jumps from joint 10th to 3rd. Legal tech teams benchmarking models for drafting work should weight groundedness over raw pass rates, and cost data matters too — Grok 4.7 tops the table at roughly US$9.50 per task, under half the ~US$21.70 of Claude Fable 5.1.

官方账号Artificial Analysis@ArtificialAnlysSource

Scaling laws for robot foundation models are far less mapped than for language models, and this 370x parameter sweep across 12 embodiments offers rare empirical evidence on that front. Using latent prediction error as a proxy for planning ability could reduce costly real-world evaluation for robotics teams. Researchers building embodied or cross-embodiment models will find the scaling trends directly comparable to their own setups.

Aran Komatsuzaki (论文推介)@arankomatsuzakiSource

This release significantly improves inference efficiency through advanced attention mechanisms and memory management, enabling faster processing of large models. Developers working with LLMs in production environments will benefit from the optimized serving capabilities and experimental features like engine snapshots, particularly for models requiring high-throughput processing.

官方账号vLLM@vllm_projectSource

NTM preserves the likelihood framework while improving few-step generation, offering a more principled approach than existing distillation or adversarial methods. Researchers and practitioners working on generative models will find this valuable for applications requiring efficient sampling with theoretical guarantees. The method provides a new direction for balancing computational efficiency with model expressiveness.

This research reveals massive inefficiency in current reinforcement learning workflows where nearly all model data is unnecessarily transferred. For developers working with large language models, NeMo-DCR offers a practical solution that could dramatically reduce training cluster costs and downtime. The approach is particularly valuable for organizations running large-scale RL experiments on distributed infrastructure like AWS.

DAIR.AI@dair_aiSource

The handbook treats agent infrastructure, knowledge governance and org design as co-equal blockers, not just model capability — a framing that shifts the AI adoption debate beyond tooling. Engineering leads and CTOs evaluating enterprise agent rollouts can use the four points as a checklist against their own readiness. The advice echoes Claude Code's Boris on building for models several months out, a useful heuristic for planning infra investment.

shao__meng@shao__mengSource