AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：arXiv:2610.20679v1 Announce Type: new Abstract: Representation-based safety defenses for vision-language models (VLMs) often learn to separate harmful from benign inputs, but separation alone does not establish what a risk score means or when it should be trusted. We introduce Counterfactual Neighborhood Risk Fields (CNRF), an evidence-grounded framework for selective VLM safety. CNRF constructs context-matched benign-harmful counterfactual pairs, uses their representation differences as capability-related risk directions, and estimates neighborhood support separately from risk. This separation prevents unfamiliarity from being conflated with harmfulness and enables selective intervention rather than blanket rejection. A victim-adaptive interface accommodates d…

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：arXiv:2610.20676v1 Announce Type: new Abstract: Correctness feedback suppresses the very evidence that specification-completeness auditing is designed to uncover: programs that satisfy a written contract but violate intended behavior. We demonstrate this objective reversal on 12 Verus/Rust tasks across four model families under matched call budgets. Independent no-feedback discovery more than doubles certified-witness yield relative to IntentShield's corrective-feedback loop (59.0% versus 27.1%). A format-matched iterative control places the larger share of this gap in feedback content, not independent restarts. The advantage holds in every model family and survives parser repair. A separate phase-separated execution matches the aggregate yield without reusing…

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：arXiv:2610.20675v1 Announce Type: new Abstract: Agents pair reasoning models with harnesses that organize model calls, execute tools, load skill instructions, and retain history in memory. Understanding agent performance requires linking execution conditions to outcomes, behavior, and resource use. We introduce AlphaDiana, a measurement system with modular interfaces for datasets, agents, environments, and scorers. It preserves harness-native control loops while controlling file reuse and memory retention separately. AlphaDiana supports macro analyses of complete agents and micro interventions on agent components such as tools, skills, and memory under matched conditions. We study three open-source models and three open-source harnesses on five answer-based ben…

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：arXiv:2610.20669v1 Announce Type: new Abstract: Reward-guided discrete diffusion enables test-time alignment with downstream objectives, but applying Sequential Monte Carlo (SMC) guidance at every denoising step is computationally expensive. Active guidance steps require additional reward evaluations and, for gradient-based proposals, value-gradient computations. Sparse guidance reduces this cost by guiding only a subset of timesteps, but commonly used schedules are fixed heuristics whose effectiveness can vary substantially across tasks. We propose VISTA, a training-free method that automatically selects guided timesteps under a fixed budget. VISTA formulates sparse guidance as approximation of the ideal full-guidance probability path and derives an estimable…

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：arXiv:2610.20667v1 Announce Type: new Abstract: Rapid advances in AI capabilities are causing benchmarks to saturate. Once several models approach perfect scores, a benchmark loses its power to differentiate between them, and with it its utility for informing AI safety cases and frontier policies. Yet benchmark items produce far more information than the scores we typically extract from them: a full transcript of how each model solved the task. We investigate whether these transcripts can be used to assess model capabilities even after benchmark scores have saturated. We simulate saturation retrospectively using historical benchmark and model releases, and test whether protocols that access only correctly solved transcripts from saturated benchmarks can predict…

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：arXiv:2610.20665v1 Announce Type: new Abstract: Large language model (LLM) agents can solve complex interactive tasks, yet reasoning and tool calls take time. In continuously changing environments, an otherwise correct action may arrive too late. We introduce RealtimeBench, with 180 tasks across six game types, to evaluate agents while games continue running during inference. Paired evaluations show that GPT-6.1-sol outperforms the faster Jev when games are paused, but falls behind when they keep running. We propose Two-Timescale Offline Recursive Self-Improvement (TORSI): a slow teacher studies failures and tests corrections in paused games between attempts, then writes reusable instructions for a fast student. The student acts without teacher calls during gam…

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Meta released Muse Spark Open, an 816B-parameter open-weight MoE model with 42B active parameters and a 1M-token context. It leads open models on MCP-Atlas and runs on 2 M5 Ultra Mac Studios. The post Meta Releases Muse Spark Open: 816B Open-Weight MoE With 1M Context appeared first on MarkTechPost.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译： Research: sparqlite: SPARQL 1.1 inside SQLite, written in Rust This research task is really an experiment in running Claude Code for web in "Ultracode" mode, as described in Orchestrate subagents at scale with dynamic workflows in the Claude Code documentation. With Ultracode turned on, Claude Code makes aggressive use of dynamic workflows - it generates JavaScript that defines and coordinates a network of subagents to take on complex problems. My initial prompt was this - I meant "hit the rows directly", the word "delete" was an accidental typo on my phone: Write a SQLite extension in Rust that implants SPARQL and runs really fast because it can hit the delete rows directly This caused Claude Code for web to churn for over 12 hours. I checked in on its progre…

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Meta says Muse excels at ‘multi-step reasoning’ and ‘complex software engineering’ as it races to catch up to Chinese rivals Meta on Thursday launched a new open-weight AI model which it claims rivals, and in some cases outperforms, similar products from leading Chinese developers. Muse Spark Open “is the most capable US open-weight model to date,” Meta’s chief of AI Alexandr Wang wrote on X. Continue reading...

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Discussion | Link

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Anant Kumar builds tools to compare loan valuations and analyze credit agreements, using Codex for code review and faster follow-ups.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Two engineers shipped Stack Overflow for Agents in a few weeks with Codex, and its ChatGPT plugin helped drive up to 10× growth in agent creation.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Reflection's first open-weight model is now available on Ollama, and free for a limited time only. Beam is designed for software engineering, reasoning, and long-running agentic tasks. Access Beam through Ollama Cloud.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：OpenAI's ChatGPT Atlas browser is gone, but now ChatGPT's Google Chrome extension is getting some upgrades that allow it to use the browser tabs you already have open. With the extension installed, OpenAI says ChatGPT can now view your browser tabs when you tag Chrome in ChatGPT's web or desktop app, posting a demo video showing someone asking ChatGPT to compare the three tabs they have open with sofas they're considering. A few big upgrades for Chrome users today! We're bringing the power of computer use to anyone who uses Chrome. First up: ChatGPT can now use your tabs in Chrome to make answers more useful and get things done for you. If … Read the full story at The Verge.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Sega is removing Total War: Shogun 2 - Complete Edition from storefronts following a disastrous launch where users complained about issues like bugs and upscaling that appears to have been done by AI. Complete Edition, a remaster of the 2011 strategy game, launched on Tuesday, and it bundled the main game and all of its DLC with promised upgrades including "improved visuals," "1440p and 4K UI scaling," and modernized controls. It has an "Overwhelmingly Negative" rating on Steam, where it has already been pulled, and posts on Reddit showcase artwork that users believe AI messed up. "We're aware that Total War: Shogun 2 - Complete Editio … Read the full story at The Verge.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Though movies about disruptive tech founders are nothing new, 2026 has been unusually stacked with features that take a much more critical look at how Silicon Valley is reshaping our reality. Aaron Sorkin's The Social Reckoning frames Mark Zuckerberg as a robotic villain whose ambitions are destabilizing society. Nathan Fielder and Lance Oppenheim's You Can See Everything paints a damning picture of Theranos founder Elizabeth Holmes as she prepares for her ongoing stint in prison. After years of being widely celebrated as a visionary, you can see how the larger narrative around Elon Musk has shifted in projects like Alex Gibney's Musk docume … Read the full story at The Verge.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：At Databricks, we continuously develop and evaluate new AI research techniques to...

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Hackers accessed roughly 1,900 GitHub accounts, then used them to slip credential-stealing workflows into more than 30,000 repositories. The malicious The post Hackers hijacked 1,900 GitHub accounts. Then they went after 30,000 repos. appeared first on The New Stack.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：President Donald Trump posted a series of photos on Truth Social showing off renovations to the White House - but one of them appears to be modified with AI. The picture, which compares an old White House column to a "restored" one, has several discrepancies and was flagged by Google's AI detector as "made or edited with OpenAI." As pointed out by users on Reddit, the picture of the "restored" column has an uneven number of steps on both sides of the structure. The number of steps also doesn't match up with the ones in the "before" photo. Additionally, the before and after photos appear nearly identical in angle and detail; the reflection i … Read the full story at The Verge.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：If your business runs on Stripe, your application needs access to Stripe data at...

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：The challenge: Two decades of logging noise at global scaleAtlassian is a global...

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：A new computational approach reveals how cells communicate to assemble into tissues. This could accelerate stem cell engineering, improve disease models, and inform therapies.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Bluesky's Attie AI tool, which initially launched earlier this year as a way to use AI to build your own feeds, now lets you make "small" social apps like games, polls, and dashboards that you can then share with others as their own websites. In an interview at TechCrunch Disrupt, Bluesky CEO Toni Schneider said to think of Attie as "social vibe coding or vibe coding for your community." You can make your own Attie app by typing in a prompt. A simple poll I made about Pokémon Red and Pokémon Blue took a few minutes to generate. If you want to check out something more full-featured, the Attie website currently highlights a few examples l … Read the full story at The Verge.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：NVIDIA released Nemotron 3.5 Super VL, an open 120B-parameter multimodal MoE model with 12B active parameters, 1M-token context and NVFP4 weights that run on 1 DGX Spark, plus 2 Bolt edge models. The post NVIDIA Releases Nemotron 3.5 Super VL, a 120B Open Multimodal Agent Model appeared first on MarkTechPost.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译： Tool: Chirp Pong Another game experiment - I wanted to see if it was feasible to have a two player game on two separate phones running in static HTML where the networking code uses chirps sent via the microphone. It seems to work! Build a two player competitive pong like game where two players use their phones close to each other and the game state is sent via audio noises which the web microphone API can interpret - but add graphical and gameplay twists to make it a lot more fun and intertwine than pong Claude Opus 5.5 Extra in Claude Code. Tags: game-design, ai, generative-ai, llms, claude-code

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Discussion | Link

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Introducing the Rosalind Academic Plan, offering eligible academic labs sponsored access to GPT-Rosalind and tools to advance life sciences research.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Vercel is giving developers more options for running their backends on its platform. At its Ship conference in San Francisco The post Vercel adds always-on workers to run more of your backend appeared first on The New Stack.

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：The new Muse Spark Open AI model can work on M5 Ultra Mac Studios. | Photo: Amelia Holowaty Krales / The Verge Meta announced the Muse Spark Open AI model today, claiming that it's the "most capable US open-weight model to date." But the company also says that it's the "the first model in its class" that can "run on your desktop using two Apple M5 Ultra Mac Studios with the latest MLX framework." Meta and Apple, typically rivals, worked together on the launch and are publicly supporting each other. "Muse Spark Open - weights releasing today - runs great on new Mac Studios thanks to work with @johnternus and Apple," Meta CEO Mark Zuckerberg said on X. "Congrats to the team!" new Apple CEO John Ternus said in a reply. "Amazing to run a model of Muse … Read the fu…

AI 服务暂时不可用，以下为来源摘要，待恢复后补全翻译：Muse Spark Open, Meta's coding-optimized open-weight model, is now available in Amazon SageMaker JumpStart. Deploy it to a SageMaker AI endpoint in your own AWS account using SageMaker Studio or the SageMaker Python SDK, and run your first inference request.