2026-10-10

Underdog Saluki 27B is a 7.89 GB, 2-bit GGUF of Qwen3.8-27B under Apache 2.0. It beats the 54 GB original on tool calling but gives up ground on competition math and reasoning. The post Meet the Underdog Saluki 27B: A 2-bit Qwen3.8-27B That Beats the Original at Tool Calling appeared first on MarkTechPost .

Speculative decoding accelerates autoregressive generation in large language models. In each drafting stage, a lightweight draft model proposes tokens that the target model subsequently verifies. With increasingly capable draft models, we find that the target model frequently accepts all tokens produced in a drafting stage. A verification nevertheless follows each drafting stage, resulting in unnecessary target-model forward passes even when drafting could have continued. Adaptive draft length m

Long-horizon agents compact their history to continue within a finite context window, but a textual summary alone may not support every subsequent decision. We introduce REMORY, a neural memory network that supplements the summary with a bounded sequence of soft memory tokens. Given the history and summary, the network learns to generate tokens that help a frozen LLM approximate the continuation it would produce with the full history. The tokens are conditioned on the summary and appended after

From the Bitter Lesson of AI scaling to the unsolved mysteries of protein folding, Google DeepMind’s Pushmeet Kohli and Biohub’s Sal Candido are rethinking what it takes to build AI that truly understands biology.

I’ve often been surprised when I hear from top researchers in industry that they think AI will be better than them at their job in a few years, and I didn’t really know why I doubted it.

联想天禧AI自主研发的专业代码智能体框架TianxiCode 以71%的问题解决率登顶全球第一名

Google Cloud has introduced the Google Cloud Gemini agent, a single agent for enterprise work. The Gemini agent is a cloud-hosted agent from Google Cloud that answers questions, does knowledge work, creates media, and writes and runs code. It does all of this from 1 prompt box and 1 API. For developers, the agent is […] The post Google Cloud Launches Gemini Agent, One Universal Agent for Enterprise Work appeared first on MarkTechPost .

arXiv:2610.11007v1 Announce Type: new Abstract: At the start of every session, LLM agents load a fixed context file, such as $\texttt{AGENTS.md}$. Each loaded token in the file is charged again in every later round of the session, and these files can degrade performance as they grow in size. However, in practice, human or automated curators usually grow these files by appending. We formulate context curation as a capacitated assortment problem. Instructions consume tokens under a finite attentio

arXiv:2610.10833v1 Announce Type: new Abstract: We study whether small LLM agents can operate effectively under explicit wall-clock time budgets by both respecting the allocated runtime and using available time productively. We evaluate Qwen3.6-27B on five competitions from MLE-Bench Lite and Qwen3-4B on Zork I (Jericho), two agentic benchmarks where additional computational time can meaningfully improve performance. In the simplest setting, where the budget is stated only in the prompt, agents

arXiv:2610.10786v1 Announce Type: new Abstract: Planning is increasingly important for long-horizon agents, where successful execution requires coordinating subgoals, tool use, and intermediate outcomes over many steps. Yet assumptions made during planning may be invalidated by the environment, tools may return unexpected results, or actions may fail. Effective agents must therefore not only generate plans, but also revise them. Such revisions often affect only part of a plan, leaving the preced

arXiv:2610.10549v1 Announce Type: new Abstract: Tool-calling agents have become central to enterprise AI, yet training and evaluating them at scale remains severely constrained due to business and legal restrictions on enterprise systems, data, and database schemas. Tabular data synthesis offers a natural alternative, but its effectiveness is fundamentally limited by structural validity and schema availability, while procedure-based approaches yield the opposite weakness, typically lacking distr

arXiv:2610.10611v1 Announce Type: new Abstract: Agentic AI systems can improve by searching longer, receiving additional support, or modifying how they propose and verify outputs. A performance score does not distinguish these mechanisms. We compare these changes through bounded verification with hidden terminal randomness. A stage specifies admissible transcripts, polynomial bounds, an alternating verification protocol, and a terminal checker. Its native reach uses default support; its closure

Explore a comprehensive coding guide to Google Research's RRSI (Regularized Recursive Self-Improvement), detailing how noise bands, cost rules, and leakage screens enable safe, efficient, and self-improving AI agents. The post Google Research RRSI Guide: Mastering Self-Improving AI Agents appeared first on MarkTechPost .

As agents take on long-horizon tasks, users shift from making individual decisions to overseeing autonomous execution. Yet the volume of agent activity and the fragmentation of supporting evidence make it difficult to determine which decisions warrant user verification. We study monitors that identify consequential decisions and locate evidence to help users assess their implications. We introduce AgentMonBench, a software-engineering benchmark comprising three subsets that cover two complementa

It seems Dario's "too powerful for you users" strategy is paying off: we have two open models at the top of the leaderboard, surpassing every single model from Anthropic. Open source prevails. Even Mistral Large 4 is better!

Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs) with limited additional computation. Beyond model scaling, this architecture has demonstrated the potential to decouple factual knowledge storage from general-purpose computation, offering a promising route to updating factual knowledge while keeping the Transformer backbone fixed. Realizing this potential is challenging because differen

Reinforcement learning from verifiable rewards (RLVR) frequently reuses rollouts across multiple policy updates, increasing the mismatch between the current policy and the data-generating policy. We identify a sign-dependent gradient starvation problem in clipped policy optimization: clipping suppresses under-generated positive responses at the low-importance-weight tail while permitting severely over-generated negative responses to dominate the high-weight tail. To address this, we propose ReSP

Spatial coding agents significantly improve spatial reasoning in Multimodal Large Language Models (MLLMs) by using external tools to generate verified execution traces. However, this paradigm inherently suffers from prohibitive inference-time overhead and external dependencies. In this paper, we explore whether an MLLM can internalize this agentic capability to operate entirely tool-free. We begin with a simple observation: prompting an MLLM with summarized execution traces of a spatial coding a

Large language models (LLMs) are increasingly used as interactive agents and simulators, yet it remains unclear whether they can recover latent sequential structure beyond surface action frequencies. This distinction is critical for behavioral simulation, where actions are often shaped by prior context rather than marginal frequencies alone. We study this question using controlled two-player Rock--Paper--Scissors interactions and a one-player stochastic n-gram continuation task. Across these exp

Blog Post : ML Drift: Next-Gen GPU AI/ML Inference at the Edge - Google Developers Blog The Google AI Edge Team is excited to announce the open-source release of ML Drift , our high-performance, cross-platform, on-device GPU compute engine specifically built for on-device AI/ML inference, under the Apache 2.0 license. By abstracting hardware and low-level API complexities of on-device GPUs across OpenGL ES, OpenCL, Metal, and WebGPU, ML Drift empowers developers to build real-time, interactive M

Long-horizon coding agents need timely corrections, yet feedback can be ineffective or even harmful when it misjudges ongoing work or fails to address the underlying problem. Existing critics focus on evaluating trajectories and generating feedback, but rarely track what happens after feedback is delivered. We present Opera, a verbal critic framework that treats each correction as a persistent note, followed until the diagnosed problem is resolved. Opera decides when to review through periodic a

(Don't judge by the screenshot, the cache is cold. It hits 24+ tok/s with a warm cache!) About two months ago, I made a post here asking whether predicting which MoE experts would be used on the next token could actually help speed up CPU/GPU offloading. Original post: Tried predicting which MoE experts get used next token to speed up CPU/GPU offload Well, quick confession first. I actually shelved that project shortly after. The reason? The speeds I was getting back then were kinda fake. My eng

I tested Mellum2.1-12B-A2.5B (Q8) locally using Pi and llama-server. All five tests were one-shot. Results were pretty mixed: - Pelican SVG, Browser OS, Minecraft: Poor results. - Bouncing Hexagon: Physics were okay, but surprisingly it made it run in the terminal. - Flappy Bird: Completed it, but the visuals were very basic and the game was way too difficult. Pelican SVG Browser OS Minecraft Flappy Bird Bouncing Hexagon Okay, these might not look great, but hear me out. This model is actually p

I got Qwen3.8-Flash-Next-GSQ-RCO-Abliterated running at IQ3_S with just 12GB VRAM and 32GB system RAM, achieving 20-30 tok/sec decode (Q2 achieves 39-45 tok/s) & 300 to ~90 thousand tok/sec prefill @ 131k context, on a custom fork of Strata. This fork has tonnes of architectural changes, all are very experimental and will probably break. But the performance makes up for it. This feels like local Opus in some regards, on sub 2k in compute.

Black-box optimization (BBO) arises in many scientific and engineering problems where objective evaluations are expensive and limited. Recent large language model (LLM) agents offer a new way to approach BBO by combining task semantics, computation, optimization tools, and feedback-driven decision making, showing great potential due to the integration with mathematically rigorous tools. However, existing agentic BBO studies use different task domains and system configurations, making their resul

I built a new feature for my blog entirely by voice with Codex Desktop, while I was cooking dinner simonwillison.net/2026/Oct/9/b...

One damaged data center has supercomputers used for training Yandex’s AI model.

Thankfully, it does not seem to have diverted police resources.

Article URL: Comments URL: Points: 44 # Comments: 8

Hello~! pocketty is an SSH terminal for iPhone and iPad, made for herdr. herdr keeps your agent panes alive on your computer and knows the state of each one: working, needs you, or done. I made this in anger/desperation for the latter half of my recent paternity leave. Nap traps are sweet, but there's only so much doom-scrolling and movie-watching I can handle... In any event, I've been using it for the last couple months and no longer have to be my desk anymore to be productive. Now the nap-tra

arXiv:2610.11005v1 Announce Type: new Abstract: Safety-aligned language models often refuse a harmful request stated directly but answer the same request inside a role-play or narrative wrapper. We measure this vulnerability across languages and registers: attack success on Qwen3-1.7B is already 89.4% in English and 93.0% in modern Chinese, and reaches 95.7% in Classical Chinese. We build GUISE, a benchmark for systematically studying this vulnerability. It includes parallel requests in English,

arXiv:2610.10942v1 Announce Type: new Abstract: Reinforcement learning environments are now a primary lever for improving large language model (LLM) capabilities in post-training, yet most agentic benchmarks remain static: the world moves only when the agent acts, the reward is a terminal verdict, and the pass bar is set arbitrarily. We introduce StoreBench, a live-commerce environment in which an agent runs a mid-size online apparel store on a production-grade commerce backend, testing long-hor

Open-source desktop AI agent for any model you choose Discussion | Link

arXiv:2610.10629v1 Announce Type: new Abstract: Self-improving LLM agents can adapt a credit pipeline to a changed rule, but an agent that rewrites itself destroys the artefact a supervisor reviews: a named change, a recorded test, an approval. We argue that self-evolution is reviewable only if it is confined to the runtime harness (instruction text, tool-call logic and primitive composition) while model weights stay fixed, so that every adaptation is a diff with a cause and a test attached. We

Find available GPUs and host open models in one command Discussion | Link

Qwen-Image-2.1-Turbo, create and edit images in just 8 denoising steps! Open weights now available! Built on Qwen-Image-2.1, Turbo is an accelerated checkpoint on the same 7B visual generation architecture. Fewer steps does not mean lower quality: it still generates strong 2K images from text, and supports continued creation through natural-language edits, from adding accessories to changing a scene. Start directly with Diffusers: load QwenImage21Pipeline and the checkpoint’s recommended 8-step

Release: ttok 1.0 I released ttok 0.4 , ran uv tool upgrade ttok , piped a file into the new version... and realized that it was defaulting to the GPT-4 tokenizer when it should very clearly default to GPT-5/GPT-6 instead! I figured switching the default was a reasonable excuse to finally ship a 1.0. OpenAI haven't actually confirmed that GPT-6 uses the same tokenizer as the GPT-5 family yet - there's an angry issue about it - but I found this commit by William Liu which reports on an experiment

Skill-augmented agents improve sample efficiency by distilling successful trajectories into reusable strategies. Yet most existing approaches remain text-centric, linearizing spatial layouts and action-state correspondences into language that loses critical geometric structure. Recent efforts have begun incorporating visual evidence, but construct and update skills separately from policy optimization, leaving their mutual improvement underexplored. We propose ViSkill, a visual-native skill learn

Most robot policies keep a model in the control loop: a VLA maps observations to actions, and an Agent Harness, such as Agent-as-Policy or Harness VLA queries a VLM for decision making at run time. We propose a different view: the embodied world is an Embodied Turing Machine, whose tape is the robot and environment state and rules are the policy. If this state can be represented accurately, the decision making can be written entirely in code. We therefore propose Code-Only-as-Policy (COAP): code

Youtu-Parsing-Omni is a compact (5B) omni-modal parsing model. Given a single input — a document page, a natural image, a chart / flowchart, a geometry figure, an audio clip or an audio-visual video — it produces one structured JSON envelope that covers both perception (layout elements, text, tables, formulas, bounding boxes, timestamps, ASR, OCR, acoustic events, camera motion) and cognition (captions, narratives, reports). The output family is selected by the task prompt ( --task in the exampl

Article URL: Comments URL: Points: 69 # Comments: 11

Batteries are now cheaper than natural gas turbines as the data center boom pushes prices up.

Debian Linux has put a call out for artist submissions for its next release: Forky.

Alibaba’s Qwen team has released Qwen-Image-2.1-Turbo, an accelerated checkpoint of its open-weight Qwen-Image-2.1 model. It generates and edits images in 8 denoising steps instead of the base model’s 40-step default. For developers, that means 5x fewer denoising steps on the same 7B architecture, plus a hosted API option. TL;DR What is Qwen-Image-2.1-Turbo? Qwen-Image-2.1-Turbo is an […] The post Alibaba Qwen Releases Qwen-Image-2.1-Turbo, an 8-Step 7B Image Model appeared first on MarkTechPost

Amazon says it will stop using NDAs when negotiating data center deals with local governments, following a similar move from Microsoft earlier this year. Secrecy has fueled community backlash against AI infrastructure, with opposition leading to hundreds of proposed and enacted moratoriums from New York to San Francisco. Meanwhile, a wave of startups is betting that consumers will hand AI agents access […]

I shipped a new feature for my blog today: the Newsletters page, which offers an index of all of the newsletters I've sent out, both my free weekly Substack and my monthly sponsors-only updates. I built the feature almost entirely using my voice, chatting away to my laptop while I cooked dinner. Codex voice mode I used the ChatGPT desktop app for this, in the Codex tab, using the voice conversation mode, running against a local development environment. Here's what that looks like: I started the

Openai just launched their decisions endpoint, cloudflare launched clef the other week, and many more jev alternatives are out there. We wanted to put the popular ones to the test and thought Pac-Man is a good benchmark for simple and fast decision making. So we let jev 1.13, kev, clef, clef flash, GPT-6 Luna and Laya play Pac-Man against bot ghosts. The low latency of these models allows for real time play. We had each model play 100 games, published a leader board and open-sourced the repo so

Everyone is very concerned about being respectable, so I’m going to be the goofball who raises worst-case possibilities. I think there is a 1% chance we live in Minicrypt, and a 15% chance we functionally lose confidence in our existing public-key encryption algorithms. [...] The problem here is that the speed of AI producing surprises, and the speed of human beings replacing standards (even with the very best AI assistance) are just orders of magnitude different. You only recover from a surpris

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. We’re putting too much faith in AI’s ability to say no Today’s AI models are trained to refuse a vast number of prompts. If you ask your chatbot how to poison…

Article URL: Comments URL: Points: 39 # Comments: 15

Three OpenAI researchers said their firings may have a 'chilling' effect on employees.

Discover how Sophos uses OpenAI’s Daybreak to cut cyber-threat investigation time by 96% and automate 52% of MDR cases while preserving human oversight.

Single-image, multi-image, and video deep research require different visual operations but share a workflow of visual grounding, external retrieval, and fact composition. A key challenge is to preserve the dependencies linking localized visual anchors, entity relations, source-supported facts, and answer-producing operations. We introduce OneSearch-VL, a unified agent centered on the Visually Grounded Evidence Graph (VGEG), which encodes these dependencies as a shared task-level reference for da

Friday, October 16, 2026 Can AI design new life forms? In 2025, Stanford University PhD student Samuel King came up with a preliminary answer when he used a generative AI model to propose genetic blueprints for microscopic viruses. It isn’t yet an example of AI-generated life, but that could be next. Join senior AI reporter…

Large Multi-modality Models (LMMs) have made significant progress in visual understanding and generation, but still face challenges in visual editing, particularly in following complex instructions, preserving appearance consistency, and supporting flexible input formats. To study this gap, we introduce RISEBench, the first benchmark for evaluating Reasoning-Informed viSual Editing (RISE), and extend it to RISEBench++, a more comprehensive and fine-grained benchmark for this emerging task. RISEB

星动纪元选择将视频预测与动作学习分阶段训练，重点不是「视频、动作一锅炖」，而是把两者「解耦」，重新「排序」。