vLLM supports Kimi K3.1 Pro on day 0: sparse attention on a KDA hybrid, a packed KV cache with FP8 or NVFP4 latents and MXFP4 index keys, token-sharded prefill, decode context parallelism, and fused low-latency decode.
arXiv:2611.01761v1 Announce Type: new Abstract: We determine the exact round complexity of Counting in anonymous directed dynamic networks. Agents have no identifiers, but $\ell\geq1$ of them are designated as leaders. Every agent knows $\ell$ and its own leader status, while the population size $n$ is unknown and has to be determined by the agents. Communication proceeds in synchronous rounds; the directed communication graph may change arbitrarily and unpredictably from one round to the next, but is assumed to remain strongly connected. We solve Counting with a distributed algorithm that makes every agent output $n$ and halt after a number of rounds equal to $\begin{cases} n&\text{if }n=\ell\\ 2n-1&\text{if }n>\ell\text{ and }\ell=1\\ 2n-2&\text{if }n>\ell\text{ and }\ell\geq2. \end{cases}$ The algorithm works even for networks with loops and parallel arcs and even when agents learn their outdegree only after sending. We prove that this bound is worst-case tight over directed multigraphs for every $n$ and $\ell$ with $1\leq\ell\leq n$. Moreover, the local computation is polynomial (provided that each communication multigraph has at most $2^{\operatorname{poly}(n)}$ arcs): in every round, each agent solves at most one linear system with up to $n$ variables. By standard techniques, solving Counting also enables the evaluation of any computable function of the multiset of agents' input values within $2n-1$ communication rounds. The bound above optimizes the worst case over all networks. For an individual network, we show that even deciding whether an agent already has enough local information to determine $n$ is $\mathsf{coNP}$-complete (provided that the numbers of parallel arcs are encoded in binary). Thus, worst-case-optimal Counting admits polynomial local computation, whereas exact network-wise optimization is intractable under standard complexity assumptions.
The Vercel CLI now includes the vercel doctor command, which diagnoses setup issues by checking CLI version and installation, login status, team affiliation, project linking, vercel.json configuration, and active incidents from the Vercel status page. It reports results as OK, Warning, Failed, or Skipped, and provides specific fix commands. The tool runs without requiring login or project linkage and supports JSON output via --json for integration with coding agents.
arXiv:2611.00036v1 Announce Type: new Abstract: Visual-prefill energy conclusions for edge vision-language models (VLMs) are commonly drawn under single-shot workloads; we revisit them under continuous perception. Profiling nine VLMs across three architecture families on Jetson AGX Thor and Orin (4,508 records, calibrated with a pure-text task), we find two mutually independent energy laws: the per-output-token decode cost grows monotonically with model scale, whereas the visual-prefill (image-encoding) cost is dominated by visual-encoder architectural efficiency, not scale. Under continuous perception ($N{=}10$) with sparse output, visual-prefill energy reaches 92--97\% of total. We derive optimization strategies---visual frame reuse, adaptive output truncation, task-aware routing, and encoder throughput optimization---for edge VLM deployment.
In distributed AI workloads, a single error can produce a cascade of symptoms. A GPU or fabric event may surface as NCCL errors on other ranks, scheduler state...
Hy4, Tencent's latest AI model with vision capabilities, is now available on Vercel's AI Gateway at 50% off through December 31, 2026. It supports reasoning, tool calling, structured outputs, image input, and a 1M-token context window. Hy4 excels in detailed image analysis, complex chart reading, and computer use across command-line and graphical interfaces. It outperforms its preview version in instruction following, long-context handling, and deep research tasks. Users can access it via the model name 'tencent/hy4' in supported tools like Claude Code, Codex, and Cursor after setting up the Vercel CLI. AI Gateway offers unified API access, usage tracking, cost management, and performance optimizations with no markup or platform fees.
Vercel now proactively checks for and blocks malicious npm packages before and during dependency installation in builds using a dynamic blocklist, preventing harmful scripts from executing. The system scans supported lockfiles and public npm downloads, automatically failing deployments if a blocked package is detected, identifying the problematic package and version, and attempting to invalidate the build cache. Users must update to a safe dependency version and regenerate the lockfile before redeploying. These security measures apply automatically to all plans without requiring configuration.
arXiv:2610.37881v1 Announce Type: new Abstract: Deepfake detection models achieve high overall accuracy but exhibit large performance disparities across demographic groups, causing some groups to experience substantially higher false-positive rates than others. Existing fairness-aware methods address this by constraining model outputs to equalize group predictions, but do not tackle the underlying feature-level bias. As a result, models may still rely on demographic characteristics, such as skin texture, as indirect cues for detecting manipulated content. To address this problem, we propose Intersectional Invariant Learning (IIL), a framework that improves feature learning rather than only adjusting model outputs. IIL alternates between identifying the most challenging demographic partitio
arXiv:2610.37680v1 Announce Type: new Abstract: Closing the gap between frontier and smaller language models is usually approached by distillation: fine-tuning the small model (student) on a frontier model's (teacher) outputs, which binds transfer to training compute and leaves the transferred knowledge largely opaque. We propose Artifact Distillation: a teacher solves training instances and externalizes its problem-solving procedure into an artifact bank that a student queries at inference time, with no gradient updates. Across four LiveBench tasks, a bank written by Claude Opus 4.5 lets Claude Haiku 4.5 outperform unaided Claude Sonnet 4.5 on spatial reasoning (0.707 vs. 0.647) at 60% of its cost, and reach Sonnet 4.5's level on table joining at 89% of its cost. The same raw bank raises
Vercel is merging its Data Cache service into Runtime Cache, unifying data caching across frameworks like Next.js, SvelteKit, and Nuxt. Developers using fetch caching or unstable_cache in Next.js won’t need to change their code. Pro and Enterprise projects receive increased storage allocations, with Pro getting 8x more space and Enterprise 10x. Runtime Cache now handles regional storage, authentication, and tag-based invalidation across Vercel’s global infrastructure. Improvements include faster reads during traffic bursts via circuit breakers, higher storage budgets reducing evictions, and enhanced observability with deployment markers, revalidation causes, and write utilization metrics. The migration simplifies billing and usage tools, with existing teams receiving advance notice and pricing details.
