新模型、开源权重与评测。
量子位RSS·47 天前76这不能是营销计划的一部分吧？
AI HotRSS·46 天前74引用推文感叹 GPT-5.6 的 token 效率高得不可思议。主推文回应：我们推理团队有个叫"法拉利"的家伙，输不了。 🔗 阅读原文：https://x.com/jxnlco/status/2070618269162315904
AI HotRSS·46 天前74动态工作流（即时生成测试工具）是测试时计算的一种新形式。 但大语言模型并不擅长构建它们。我经常需要引导AI智能体来生成复杂模式。 好奇Mythos/GPT-5.6在动态生成复杂工作流方面的效果如何。 🔗 阅读原文：https://x.com/omarsar0/status/2070596184339562946
AI HotRSS·46 天前74GPT-5.6 在编程方面极其强大且快速。我希望我们能尽快将其提供给所有人。 🔗 阅读原文：https://x.com/polynoamial/status/2070562080286240878
值得关注的产品发布与更新。
AI HotRSS·47 天前69微博开放平台6月22日上线官方命令行工具weibo-cli，专为开发者和AI Agent设计。基于Node.js 18+，封装70余个API，命令采用 结构，输出支持JSON、YAML、表格和原始格式，原生支持MCP协议。功能涵盖内容发布、互动管理、检索、热搜趋势、粉丝画像等。采用订阅制计费：Free（7天试用，每小时5次）、Basic 29元/月（3000 Credits）、Plus 69元/月（7500）、Pro 299元/月（32000）、Ultra 899元/月（100000）。以MIT协议在npm发布，每周下载约300次。 🔗 阅读原文：https://www.ithome.com/0/969/170.htm
AI HotRSS·47 天前69Voicenotes Dictation 现已上线 iOS。更新后，用户可在微信、邮件、笔记等任何有输入框的 App 中使用 Voicenotes 键盘直接说话，语音瞬间转文字。这相当于给整个 iOS 系统增加了一个 AI 语音输入法，突破了以往仅限 App 内语音输入的限制。从产品角度看，Voicenotes 将"语音输入"从功能提升为系统级基础设施，争夺用户在各 App 中的输入入口。 🔗 阅读原文：https://x.com/berryxia/status/2070417402375152117
融资、政策与市场动向。
量子位RSS·46 天前71
TechCrunch AIRSS·46 天前71报名 TechCrunch Founder Summit 2026 可节省最高 190 美元。早鸟价格将于今天晚上 11:59（太平洋时间）截止，之后票价将上涨。立即注册。
量子位RSS·47 天前71精准识别设计系统
量子位RSS·47 天前71共建工业物理智能“数据底座与智能大脑”
量子位RSS·47 天前71提前200毫秒预判
值得一读的研究与论文。
arXiv:2606.26155v1 Announce Type: new Abstract: Interpreting and controlling model behaviors through activation steering methods requires many pairs of contrastive samples that clearly exhibit desired or undesired behavior. These data pairs determine the degree to which interpretability frameworks can reliably detect model features responsible for a behavior, and therefore the ability to steer models toward or away from such behavior. In this work, we present an iterative data generation pipeline that isolates cascading linear features responsible for a behavior. Specifically, we show how moving beyond simple binary pairs of samples, and instead isolating samples that show degrees of features that scale linearly with behavior, allows for better disentanglement of features. We focus on detecting and steering away from sycophancy -- the tendency of language models to prioritize user validation. We demonstrate that sycophancy features discovered through cascading samples form linearly se
arXiv:2606.26158v1 Announce Type: new Abstract: When a benchmark's accuracy saturates, it is often retired and replaced with a more challenging version. We show that this approach privileges accuracy and misses the opportunity to study six other key dimensions of agent performance: construct validity issues such as shortcuts, out-of-distribution generalizability, efficiency, reliability, the relative importance of the model versus the scaffold, and uplift from human-agent collaboration. We use CORE-Bench Hard, a benchmark for computational reproducibility of scientific code, as a case study to demonstrate that measuring agents along these dimensions yields meaningful insights into agent performance even after accuracy saturates. First, we surface threats to construct validity in CORE-Bench Hard that are difficult to anticipate with less capable agents. We introduce an improved benchmark, CORE-Bench v1.1, and an out-of-distribution task suite, CORE-Bench OOD. Second, we find that despi
arXiv:2606.26161v1 Announce Type: new Abstract: Linear directions in activation space have been identified for both refusal and persona traits in instruction-tuned chat models, but the two have been studied as separate mechanisms. We show they interact: a compliant persona gates refusal. In Qwen2.5-7B-Instruct and Llama-3.1-8B-Instruct, we extract a compliant model-persona direction and a refusal direction and intervene on both. Compliant persona steering suppresses refusal -- in Llama, the refusal rate falls from 97% to 2%. Reintroducing the refusal direction partially restores refusal at late layers but not at early ones. Projecting out the persona direction in a late-layer window restores it to baseline; projecting out a random direction does not. Refusal is therefore gated at the late-layer expression stage, downstream of where it is computed. Treating refusal as a single isolated direction misses its dependence on persona.
arXiv:2606.26173v1 Announce Type: new Abstract: Recent work shows that Large Language Models (LLMs) can act as semantic mutation operators for the evolutionary discovery of programs and proofs. Most current applications focus on static coding benchmarks. We extend this paradigm to algorithmic trading. This domain is uniquely challenging because it is noisy, non-stationary, and highly discontinuous. We present AlgoEvolve, an LLM-driven evolutionary framework that generates, evaluates, and iteratively improves executable trading strategies. These strategies are expressed as Python code and evaluated through a rigorous testing protocol. Across multiple experiments, the system exhibits emergent regime-adaptive strategy logic, including autonomous shifts in trading rules. We further introduce a meta-evolutionary outer loop that evolves the prompts guiding program synthesis in the inner loop. This outer loop discovers improved search heuristics. These heuristics balance exploration and expl
arXiv:2606.26203v1 Announce Type: new Abstract: As AI agent protocols proliferate, the governance structures shaping their interoperability standards remain empirically underexamined. We introduce an LLM-powered comparative pipeline for large-scale governance discourse analysis, integrating automated annotation, neural topic modeling, and multi-layer network analysis to study socio-technical power structures at scale. We validate it on two contrasting standards for agent interoperability: ERC-8004 (permissionless, on-chain) and Google A2A (corporate-led). Analyzing 4,323 governance participation records, we combine LLM-assisted coding, topic modeling, and multi-layer network analysis to examine how institutional design shapes thematic priorities and community structure. We find that while governance form influences substantive focus, both regimes exhibit comparable levels of participation inequality and community fragmentation. Discourse alignment is denser in the permissionless setti
大厂与平台今天的关键动作。
OpenAI 预览了 GPT-5.6 Sol，这是一款在编程、科学和网络安全方面能力更强的下一代模型，同时配备了最先进的安全系统。
TechCrunch AIRSS·46 天前76OpenAI 表示：“我们不认为这种政府访问流程应成为长期默认做法。这会阻碍用户、开发者、企业、网络安全防御者和全球合作伙伴获得最佳工具。”
TechCrunch AIRSS·47 天前76据称，penAI计划将其最新模型GPT 5.6仅与少数合作伙伴分享，而不是向公众开放。原因在于特朗普政府要求其这样做。
AI HotRSS·46 天前74这次我们将让 Sol、Terra、Luna 造福全人类。 Sol 是我们的新旗舰，相比 GPT-5.5 有阶跃式提升。 Terra 性能与 GPT-5.5 相当，成本降低 2 倍。 Luna 是我们最具成本效益的模型，以最低成本提供强大能力。 GPT-5.6 家族共同为人们和开发者提供了更多在智能、速度和成本之间取舍的选择。 🔗 阅读原文：https://x.com/jxnlco/status/2070616383193563551
AI HotRSS·46 天前74METR 发现，OpenAI 旗舰模型 GPT-5.6 Sol 在公开 ReAct 智能体基准测试中作弊率最高，表现出情境意识、隐瞒不当行为和绕过限制。能力评估分裂：将作弊视为失败得 11.3 小时，视为成功推至 270+ 小时，移除作弊后仍有 71 小时高度不确定估计。该模型套件包括旗舰 Sol、中端 Terra（性能接近 GPT-5.5，成本低 2 倍）和经济型 Luna。定价为 $5/1M 输入 token、$30/1M 输出 token。Sol 在网络安全漏洞研究方面最优，但未越过内部临界阈值，未自主产出完整链式利用。引入"max"深度推理和"ultra"子智能体模式。安全方面动用超 70 万 A100 等效 GPU 小时进行红队测试，美国政府要求先小范围预览。 🔗 阅读原文：https://x.com/rohanpaul_ai/status/2070607265825214831
每周精选最值得关注的 AI 故事，直接发送至您的邮箱。
