Search AI news

BlockLock: Reproducible LLM Inference with Fixed Batch Geometry

arXiv:2611.06977v1 Announce Type: new Abstract: Reproducible LLM inference requires that a request produces the same output regardless of the other requests in the batch. Existing solutions pay for this in different ways: batch-invariant kernels constrain every operator, and isolating prefill with fixed-shape decode ({our CoRun-style baseline}) gives up shared prefill compute and pads decode. We present BlockLock, which keeps shared batched prefill and pins only the numerical path of the protected request: it fixes the geometry that the request sees, with {two modes: BlockLock-Rows uses fixed-width rows and retains batched decode of all requests; BlockLock-Pack uses terminal-anchored packing with an isolated dummy segment, then extracts the protected request's KV state for target-only decode. Under a stated per-deployment contract, both modes preserve the target's output bitwise when its companions change.} Across five {dense} models (Qwen3-8B/14B, Llama-3.1-8B-Instruct, and Gemma-3-4B/12B-IT) and three single-GPU deployments (NVIDIA A6000, A100, and H100), we evaluate our approach on both tinyBenchmarks and BFCL{; we also include initial tensor-parallel experiments on two NVIDIA A6000 GPUs. Both modes preserve target sequences and Top20-logit bits in the reported tests.} At equal delivered work, {BlockLock-Rows} reduces latency by 53.5--55.0\% and improves throughput by 115.2--122.2\% relative to vLLM Batch-Invariant; compared with batch-invariant kernels, it needs no custom kernel implementations and constrains only the protected requests.

FlatCake: A Flat KVCache Storage Architecture for LLM Serving

arXiv:2610.38355v1 Announce Type: new Abstract: KV cache reuse accelerates LLM serving, but existing systems manage host DRAM and NVMe SSD as a tiered hierarchy where DRAM acts as an intermediate cache. Production LLM workloads break both premises of this hierarchy. First, bimodal reuse defeats DRAM temporal filtering, since workloads are dominated either by one-shot blocks with long reuse distances or by multi-gigabyte agentic sessions that exceed DRAM capacity, putting SSD on the critical read path. Second, DRAM staging and eviction-ordered layout bottleneck SSD restoration. Holding host staging buffers for in-flight prefixes causes contention under load that aborts prefetches, while incremental eviction scatters a session's blocks across the SSD and prevents large sequential reads. We p

Prefix Keys and Values Matter Late: RIPE Caching for Self-Speculative Decoding in Looped Language Models

arXiv:2610.34436v1 Announce Type: new Abstract: A looped language model runs one block of layers $R$ times per token, so its key-value (KV) cache holds $R$ planes per recurrent layer, and self-speculative decoding (draft with a few loops, verify with all $R$) reads them all. We find that prefix keys and values matter late: at equal perturbation size, loops 25-32 are $e^{15.0}$ (Huginn-0125) and $e^{12.8}$ (Raven-Llama) times more sensitive than loops 2-9, and the planes approach the last one geometrically, at 0.88 and 0.81 per loop. Whether the deepest plane alone suffices then depends on how ripe the late planes are: on Raven, storing only it is indistinguishable from a bf16-to-fp32 change; on Huginn, 1% of positions carry 74% of its error. We turn this into RIPE, which stores the deepest

Democratizing LLM Training with the Edge: Challenges, Systems, and the Road Ahead

arXiv:2610.32215v1 Announce Type: new Abstract: The growing computational power of modern GPU data centers has let a few industrial giants train and deploy large language models (LLMs) on massive amounts of data. Yet the sheer scale of these models and their training data makes centralized development increasingly hard to sustain, and at the same time shuts out smaller institutions that hold relevant private data but lack comparable infrastructure. In this paper, we organize our positions on distributed LLM training around the pretraining and fine-tuning paradigms, covering federated learning, split learning, and decentralized approaches, and we assess their readiness for geo-distributed settings with heterogeneous hardware and unreliable connectivity. We then present a {\em unified} syste

Nanbeige Releases Nanbeige4.3-33B, a 33B Open Agentic Coding Model

Nanbeige LLM Lab released Nanbeige4.3-33B, a 33B dense Apache 2.0 model with 512K context. It reports 80.9 on Terminal-Bench 2.1 and will teach the upcoming Nanbeige4.3-3B. The post Nanbeige Releases Nanbeige4.3-33B, a 33B Open Agentic Coding Model appeared first on MarkTechPost.

Hugging Face Releases Carbon-X, a 14B Open DNA Language Model that Brings DNA and Natural Language Together for the First Time

Hugging Face released Carbon-X, an open 14.8B DNA language model built on Qwen3-14B. It reads raw DNA in chat, answers in English, and averages 66.7% across 5 DNA benchmarks under Apache 2.0. The post Hugging Face Releases Carbon-X, a 14B Open DNA Language Model that Brings DNA and Natural Language Together for the First Time appeared first on MarkTechPost.

Research reveals the local geometric structure of semantic meaning in large language models

A paper published on October 24, 2026, on arXiv (ID 2610.30715v1) characterizes the local geometric properties of semantic meaning in embedding space across four decoder-only large language models—Gemma 4, Llama 3.2, Qwen 2.5, and Pythia—using cross-lingual probes and fine-grained probes.

Why it matters：This study is the first to systematically characterize the invariance, variability, and refinement mechanisms within “the same semantic factor” in LLMs, proposing the “local anisotropic neighborhood” model. The approach has been validated through blind human evaluations, demonstrating rigorous methodology, broad model coverage, and cross-architecture consistency in its findings.

Research on Modulating the Capability Distribution of Large Language Models via Singular Value Spectrum Intervention: Empirical Analysis of Llama and Qwen Models

An AI study published on October 24, 2026, on arXiv (ID 2610.30676v1) applied post-hoc singular-value spectrum smoothing and spike perturbations to five open-source Llama and Qwen instruction-tuned models, finding that this lightweight intervention can systematically reshape the trade-offs in model performance across three capability domains: factual knowledge, language understanding, and deductive reasoning.

Why it matters：This paper is the first to empirically demonstrate, within a multi‑model, multi‑task framework, the causal impact of the singular value spectrum of transformer feedforward layers on downstream capabilities, revealing structurally sensitive patterns in capability allocation and offering a new pathway for decoupling model capabilities and enabling controllable editing. All conclusions are based on publicly available models and standard benchmarks, with transparent methods and comprehensive evaluation dimensions.

OpenAI's latest GPT-6 series models will be available on Google Cloud and the Antigravity platform.

OpenAI’s cutting-edge GPT-6 and 6.1 series models (Astra, Sol, Luna) will be launched on Google Cloud Model Garden in the coming weeks and integrated into Google’s agent-driven programming tool, Antigravity; this move makes Google the first mainstream cloud provider to offer all three major closed-source large models—Gemini, Claude, and GPT.

FairStress: A Novel Fairness Evaluation Framework for Symmetric Robustness in LLM Decision-Making

FairStress is a new benchmark that focuses on the decision stability, interpretability consistency, and internal representation symmetry of large language models (LLMs) in demographic identity substitution scenarios. It reveals systematic behavioral biases in these models, such as “anti-stereotypical over-correction,” and highlights that mainstream evaluation methods—like BBQ—may misjudge the direction of bias due to paradigmatic differences.

Qwen2.5-Max: Exploring the Intelligence of Large-scale MoE Model

QWEN CHAT API DEMO DISCORD It is widely recognized that continuously scaling both data size and model size can lead to significant improvements in model intelligence. However, the research and industry community has limited experience in effectively scaling extremely large models, whether they are dense or Mixture-of-Expert (MoE) models. Many critical details regarding this scaling process were only disclosed with the recent release of DeepSeek V3. Concurrently, we are developing Qwen2.

Qwen2.5: A Party of Foundation Models!

GITHUB HUGGING FACE MODELSCOPE DEMO DISCORD Introduction In the past three months since Qwen2’s release, numerous developers have built new models on the Qwen2 language models, providing us with valuable feedback. During this period, we have focused on creating smarter and more knowledgeable language models. Today, we are excited to introduce the latest addition to the Qwen family: Qwen2.5. We are announcing what might be the largest opensource release in history!

Qwen2.5-LLM: Extending the boundary of LLMs

GITHUB HUGGING FACE MODELSCOPE DEMO DISCORD Introduction In this blog, we delve into the details of our latest Qwen2.5 series language models. We have developed a range of decoder-only dense models, with seven of them open-sourced, spanning from 0.5B to 72B parameters. Our research indicates a significant interest among users in models within the 10-30B range for production use, as well as 3B models for mobile applications. To meet these demands, we are open-sourcing Qwen2.

Qwen1.5-MoE: Matching 7B Model Performance with 1/3 Activated Parameters

GITHUB HUGGING FACE MODELSCOPE DEMO DISCORD Introduction Since the surge in interest sparked by Mixtral, research on mixture-of-expert (MoE) models has gained significant momentum. Both researchers and practitioners are keenly interested in understanding how to effectively train such models and assessing their efficiency and effectiveness. Today, we introduce Qwen1.5-MoE-A2.7B, a small MoE model with only 2.7 billion activated parameters yet matching the performance of state-of-the-art 7B models like Mistral 7B and Qwen1.

Introducing Qwen1.5

GITHUB HUGGING FACE MODELSCOPE DEMO DISCORD Introduction In recent months, our focus has been on developing a “good” model while optimizing the developer experience. As we progress towards Qwen1.5, the next iteration in our Qwen series, this update arrives just before the Chinese New Year. With Qwen1.5, we are open-sourcing base and chat models across six sizes: 0.5B, 1.8B, 4B, 7B, 14B, 32B, 72B, and 110B, and also an MoE model (see blog for more information).