每日 AI 情报报告

生成时间：2026-03-26T07:02:52+08:00（Asia/Shanghai）

抓取 20 个来源，原始 359 条，清洗后 63 条，去重后 58 条。

固定链接：https://larry-openclaw-reports.pages.dev/reports/ai-daily/

摘要（国外 / 国内）

国际重点 AI 新闻（优先）

海外媒体、官方实验室与前沿论文信源默认置顶，国内来源统一放在报告下方作为补充。

• Reliable Classroom AI via Neuro-Symbolic Multimodal Reasoning 重点国际前沿行业趋势普通质量分 64.7 · 前沿分 79.4 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22793v1 Announce Type: new Abstract: Classroom AI is rapidly expanding from low-level perception toward higher-level judgments about engagement, confusion, collaboration, and instructional quality. Yet classrooms are among the hardest real-world se…

  apibenchmarkinference
• Memory Bear AI Memory Science Engine for Multimodal Affective Intelligence: A Technical Report 重点国际前沿开源项目持续热议质量分 72.1 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22306v1 Announce Type: new Abstract: Affective judgment in real interaction is rarely a purely local prediction problem. Emotional meaning often depends on prior trajectory, accumulated context, and multimodal evidence that may be weak, noisy, or i…

  benchmarkinference
• Bridging the Know-Act Gap via Task-Level Autoregressive Reasoning 重点国际前沿模型发布普通质量分 69.8 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22619v1 Announce Type: new Abstract: LLMs often generate seemingly valid answers to flawed or ill-posed inputs. This is not due to missing knowledge: under discriminative prompting, the same models can mostly identify such issues, yet fail to refle…

  benchmarkllm
• Improving Safety Alignment via Balanced Direct Preference Optimization 重点国际前沿模型发布普通质量分 66.7 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22829v1 Announce Type: new Abstract: With the rapid development and widespread application of Large Language Models (LLMs), their potential safety risks have attracted widespread attention. Reinforcement Learning from Human Feedback (RLHF) has been…

  apibenchmarkllmtraining
• Dynamical Systems Theory Behind a Hierarchical Reasoning Model 重点国际前沿模型发布普通质量分 64.5 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22871v1 Announce Type: new Abstract: Current large language models (LLMs) primarily rely on linear sequence generation and massive parameter counts, yet they severely struggle with complex algorithmic reasoning. While recent reasoning architectures…

  benchmarkllmtraining
• Session Risk Memory (SRM): Temporal Authorization for Deterministic Pre-Execution Safety Gates 重点国际Agent/自动化持续热议质量分 71.9 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22350v1 Announce Type: new Abstract: Deterministic pre-execution safety gates evaluate whether individual agent actions are compatible with their assigned roles. While effective at per-action authorization, these systems are structurally blind to d…

  agentbenchmarkinferencetraining
• Maximum Entropy Relaxation of Multi-Way Cardinality Constraints for Synthetic Population Generation 重点国际模型发布持续热议质量分 70.9 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22558v1 Announce Type: new Abstract: Generating synthetic populations from aggregate statistics is a core component of microsimulation, agent-based modeling, policy analysis, and privacy-preserving data release. Beyond classical census marginals, m…

  agentbenchmarkrelease
• Benchmarking Multi-Agent LLM Architectures for Financial Document Processing: A Comparative Study of Orchestration Patterns, Cost-Accuracy Tradeoffs and Production Scaling Strategies 重点国际模型发布普通质量分 69.9 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22651v1 Announce Type: new Abstract: The adoption of large language models (LLMs) for structured information extraction from financial documents has accelerated rapidly, yet production deployments face fundamental architectural decisions with limit…

  agentapibenchmarkllm
• Can LLM Agents Generate Real-World Evidence? Evaluating Observational Studies in Medical Databases 国际开源项目普通质量分 66.5 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22767v1 Announce Type: new Abstract: Observational studies can yield clinically actionable evidence at scale, but executing them on real-world databases is open-ended and requires coherent decisions across cohort construction, analysis, and reporti…

  agentapibenchmarkllm
• Dynamic Fusion-Aware Graph Convolutional Neural Network for Multimodal Emotion Recognition in Conversations 国际其他重要情报持续热议质量分 64.3 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22345v1 Announce Type: new Abstract: Multimodal emotion recognition in conversations (MERC) aims to identify and understand the emotions expressed by speakers during utterance interaction from multiple modalities (e.g., text, audio, images, etc.). …

  inference
• Empirical Comparison of Agent Communication Protocols for Task Orchestration 国际Agent/自动化普通质量分 63.5 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22823v1 Announce Type: new Abstract: Context. Nowadays, artificial intelligence agent systems are transforming from single-tool interactions to complex multi-agent orchestrations. As a result, two competing communication protocols have emerged: a t…

  agentbenchmark
• Chain-of-Authorization: Internalizing Authorization into Large Language Models via Reasoning Trajectories 国际模型发布普通质量分 63.3 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22869v1 Announce Type: new Abstract: Large Language Models (LLMs) have become core cognitive components in modern artificial intelligence (AI) systems, combining internal knowledge with external context to perform complex tasks. However, LLMs typic…

  llmtraining

Builder 雷达（X / YouTube / GitHub）

重点覆盖 AI 公司高管、技术负责人、硅谷 builder、AI 高级工程师与 team leader。

X 重点人物动态 (40)

• Claude (@claudeai)Claude is an AI assistant built by @anthropicai to be safe, accurate, and secure. Talk to Claude on https://t.co/ZhTwG8… · 互动 37728（👍31516 / 转2235 / 评1742） · 2026-03-24T18:01:05.000Z

  New in Claude Code: auto mode. Instead of approving every file write and bash command, or skipping permissions entirely, auto mode lets Claude make permission decisions on your behalf. Safeguards check each action before it runs. https://t…

• Andrej Karpathy (@karpathy)I like to train large deep neural nets. Previously Director of AI @ Tesla, founding team @ OpenAI, PhD @ Stanford. · 互动 30187（👍21171 / 转4006 / 评1004） · 2026-03-24T16:56:24.000Z

  Software horror: litellm PyPI supply chain attack. Simple `pip install litellm` was enough to exfiltrate SSH keys, AWS/GCP/Azure creds, Kubernetes configs, git credentials, env vars (all your API keys), shell history, crypto wallets, SSL p…

• Sam Altman (@sama)AI is cool i guess · 互动 7721（👍5437 / 转449 / 评1386） · 2026-03-24T17:01:52.000Z

  AI will help discover new science, such as cures for diseases, which is perhaps the most important way to increase quality of life long-term. AI will also present new threats to society that we have to address. No company can sufficiently …

• Alex Albert (@alexalbert__)Research @AnthropicAI. Opinions are my own! · 互动 2224（👍1972 / 转72 / 评108） · 2026-03-24T18:27:24.000Z

  Goodbye --dangerously-skip-permissions, hello auto mode https://t.co/1r2nUd3O7P

• Guillermo Rauch (@rauchg)@vercel CEO · 互动 2104（👍1654 / 转119 / 评212） · 2026-03-24T14:19:44.000Z

  Almost every SaaS app inside Vercel has now been replaced with a generated app or agent interface, deployed on Vercel. Support, sales, marketing, PM, HR, dataviz, even design and video workflows. It’s shocking. The SaaSpocalypse is both un…

• Josh Woodward (@joshwoodward)VP, @Google @GoogleLabs @GeminiApp @GoogleAIStudio · 互动 1768（👍1483 / 转103 / 评79） · 2026-03-24T18:38:32.000Z

  I've been at @Google since I was an intern, and there's never been a more exciting time. The place is pulsating. We're hiring :) @GeminiApp or @GoogleAIStudio: https://t.co/otVHg9uuqo @GoogleLabs: https://t.co/e7zA6GJYDe https://t.co/9J4UF…

• Claude (@claudeai)Claude is an AI assistant built by @anthropicai to be safe, accurate, and secure. Talk to Claude on https://t.co/ZhTwG8… · 互动 1648（👍1544 / 转29 / 评46） · 2026-03-24T18:01:05.000Z

  Before each tool call, a classifier reviews it for potentially destructive actions. Safe actions proceed automatically. Risky ones get blocked, and Claude takes a different approach. This reduces risk but doesn't eliminate it. We recommend…

• Thariq (@trq212)Claude Code @anthropicai. prev YC W20, mit media lab. towards machines of loving grace · 互动 1366（👍1180 / 转35 / 评116） · 2026-03-24T18:38:39.000Z

  turns out being an AI safety company is useful for when you need to make sure AIs can run safely https://t.co/DjJZjPrZn4

• Claude (@claudeai)Claude is an AI assistant built by @anthropicai to be safe, accurate, and secure. Talk to Claude on https://t.co/ZhTwG8… · 互动 1220（👍1085 / 转39 / 评57） · 2026-03-24T18:01:05.000Z

  Available now as a research preview on the Team plan. Enterprise and API access rolling out in the coming days. Enable with claude --enable-auto-mode, then cycle to it with Shift+Tab. Learn more: https://t.co/eOurVAnoKc

• Peter Steinberger (@steipete)Polyagentmorous ClawFather. Came back from retirement to mess with AI and help a lobster take over the world @openclaw🦞 · 互动 927（👍835 / 转19 / 评54） · 2026-03-24T16:52:37.000Z

  Messaged a friend on WhatsApp, got greeted by the OpenClaw doorman 🤣 (use separate number + WA Business to avoid this, see the docs, or use a message platform that is friendlier for agents such as Telegram) https://t.co/9iVu6lb1tf

• Amjad Masad (@amasad)ceo @replit. civilizationist · 互动 648（👍552 / 转36 / 评24） · 2026-03-24T12:57:50.000Z

  If you’re a content creator you need to watch this. No hype or marketing. Nick is sharing is his genuine journey on how learning to make software transformed their business. https://t.co/YiZ89b36PW

• Sam Altman (@sama)AI is cool i guess · 互动 511（👍393 / 转32 / 评54） · 2026-03-24T17:02:03.000Z

  More details here: https://t.co/OT0cPOCGO4

• Nikunj Kothari (@nikunj)partner @fpvventures - investing in seed/A. previous: early hire @meter, @opendoor, @atlassian & others. love @shimolee… · 互动 478（👍394 / 转15 / 评54） · 2026-03-24T22:33:50.000Z

  SOC II is in the news right now for being security theater.. You know what SOC II is *actually* good for? Subprocessor lists. I scraped 417 companies subprocessors to investigate what AI native companies are using for their infrastructure.…

• Aaron Levie (@levie)ceo @box - your business lives in content. unleash it with AI · 互动 475（👍361 / 转31 / 评52） · 2026-03-24T16:23:23.000Z

  “We’ve also been moving off legacy systems with poor, slow, outdated, and inconsistent APIs.” If you’re building software that can’t work fully headlessly in a way that agents want to use, you’re not prepared for what the future of softwar…

• Zara Zhang (@zarazhangrui)Builder. Dangerously skips permissions. Harvard’17. GitHub: https://t.co/KCuEajezlL YouTube: https://t.co/8xzbGWtf6w · 互动 438（👍377 / 转12 / 评37） · 2026-03-24T23:47:20.000Z

  Something shifted in January... Last year I barely knew how GitHub worked; now I have 13k+ stars on GitHub (and I'm not even a technical person, but then again how do you define technical these days?) https://t.co/tpqvp2eanG

• Nan Yu (@thenanyu)head of product @linear · 互动 297（👍266 / 转11 / 评9） · 2026-03-24T21:04:07.000Z

  I haven’t written a PRD by hand, filed an issue through a form, or hand-written any code in months. But the volume of work I’m producing and the quality bar have never been higher. https://t.co/dSyyn7yzwg

• Cat Wu (@_catwu)claude code + cowork @anthropicai, prev: @dagster, @scale_ai · 互动 259（👍211 / 转10 / 评28） · 2026-03-25T00:02:56.000Z

  Excited to see everyone at Code with Claude! What would you like to hear from us at this year's sessions? https://t.co/RxWLfs4kFo

• Thariq (@trq212)Claude Code @anthropicai. prev YC W20, mit media lab. towards machines of loving grace · 互动 225（👍190 / 转10 / 评15） · 2026-03-24T13:59:56.000Z

  I’ll be covering how to make the most of this in my livestream on March 31st with Figma! You can sign up here: https://t.co/G32hMTFUR4 https://t.co/9owzK6bjWF

• Guillermo Rauch (@rauchg)@vercel CEO · 互动 171（👍153 / 转2 / 评14） · 2026-03-24T23:24:05.000Z

  Vercel can now intelligently pick the right hardware for your build. With new Rust-based compilers like Turbopack & Rolldown, build performance now scales with 𝒪(cpus). But too many CPUs and you waste money. Too few and agents waste time. …

• Swyx (@swyx)achieve ambition with intentionality, intensity, integrity & insanity. affiliations: - @dxtipshq - @cognition - @tempor… · 互动 148（👍122 / 转1 / 评24） · 2026-03-24T20:00:33.000Z

  first casualty of OpenAI’s crackdown on Side Quests: Sora is dead :( https://t.co/sfQhNu3IQ5

X 监控账号池 (39)

• Aaron Levie (@levie)、Aditya Agarwal (@adityaag)、Alex Albert (@alexalbert__)、Amanda Askell (@AmandaAskell)、Amjad Masad (@amasad)、Andrej Karpathy (@karpathy)、Andrew Ng (@AndrewYNg)、Aravind Srinivas (@AravSrinivas)、Cat Wu (@_catwu)、Claude (@claudeai)、Dan Shipper (@danshipper)、Dario Amodei (@DarioAmodei)、Demis Hassabis (@demishassabis)、Elon Musk (@elonmusk)、Fei-Fei Li (@drfeifei)、Francois Chollet (@fchollet)、Garry Tan (@garrytan)、Google Labs (@GoogleLabs)、Greg Brockman (@gdb)、Guillermo Rauch (@rauchg)、Jeff Dean (@JeffDean)、Josh Woodward (@joshwoodward)、Kevin Scott (@kevin_scott)、Kevin Weil (@kevinweil)、Madhu Guru (@realmadhuguru)、Mark Zuckerberg (@finkd)、Matt Turck (@mattturck)、Nan Yu (@thenanyu)、Nikunj Kothari (@nikunj)、Peter Steinberger (@steipete)、Peter Yang (@petergyang)、Ryo Lu (@ryolu_)、Sam Altman (@sama)、Satya Nadella (@satyanadella)、Sundar Pichai (@sundarpichai)、Swyx (@swyx)、Thariq (@trq212)、Yann LeCun (@ylecun)、Zara Zhang (@zarazhangrui)

YouTube AI 播客动态 (1)

• Latent Space: 🔬There Is No AlphaFold for Materials — AI for Materials Discovery with Heather Kulik2026-03-24T00:00:00.000Z

  There's a school of thought that why should I bother to learn chemistry or physics or whatever when TAGPT know has PhD level understanding of that. Anyway, [music] >> Chach GPT is super good at Wikipedia level chemistry knowledge. I'm real…

YouTube 监控频道池 (5)

• Data Driven NYC (youtube_channel)、Latent Space (youtube_channel)、No Priors (youtube_channel)、Training Data (youtube_playlist)、Unsupervised Learning (youtube_channel)

GitHub 星标增长（日榜） (9)

• bytedance/deer-flow今日 +3787 星标 · 总星标 46070 · Python

  Star bytedance / deer-flow An open-source SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours.

• Crosstalk-Solutions/project-nomad今日 +1717 星标 · 总星标 16560 · TypeScript

  Star Crosstalk-Solutions / project-nomad Project N.O.M.A.D, is a self-contained, offline survival computer packed with critical tools, knowledge, and AI to keep you informed and empowered—anytime, anywhere.

• mvanhorn/last30days-skill今日 +1342 星标 · 总星标 7509 · Python

  Star mvanhorn / last30days-skill AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary

• ruvnet/ruflo今日 +1173 星标 · 总星标 26132 · TypeScript

  Star ruvnet / ruflo 🌊 The leading agent orchestration platform for Claude. Deploy intelligent multi-agent swarms, coordinate autonomous workflows, and build conversational AI systems. Features enterprise-grade architecture, distributed swarm intelligence, RAG integration, and native Claude Code / Codex Integration

• supermemoryai/supermemory今日 +809 星标 · 总星标 19178 · TypeScript

  Star supermemoryai / supermemory Memory engine and app that is extremely fast, scalable. The Memory API for the AI era.

• hsliuping/TradingAgents-CN今日 +449 星标 · 总星标 21288 · Python

  Star hsliuping / TradingAgents-CN 基于多智能体LLM的中文金融交易框架 - TradingAgents中文增强版

• BerriAI/litellm今日 +289 星标 · 总星标 40622 · Python

  Sponsor Star BerriAI / litellm Python SDK, Proxy Server (AI Gateway) to call 100+ LLM APIs in OpenAI (or native) format, with cost tracking, guardrails, loadbalancing and logging. [Bedrock, Azure, OpenAI, VertexAI, Cohere, Anthropic, Sagemaker, HuggingFace, VLLM, NVIDIA NIM]

• usestrix/strix今日 +215 星标 · 总星标 21703 · Python

  Star usestrix / strix Open-source AI hackers to find and fix your app’s vulnerabilities.

• letta-ai/claude-subconscious今日 +71 星标 · 总星标 1421 · TypeScript

  Star letta-ai / claude-subconscious Give Claude Code a subconscious

分类情报区

按分类浏览，分类内按质量分与时间排序。

AI模型与产品发布 (0)

• 该分类暂无数据

大模型行业动态 (0)

• 该分类暂无数据

模型发布 (23)

• Bridging the Know-Act Gap via Task-Level Autoregressive Reasoning 重点国际前沿模型发布普通质量分 69.8 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22619v1 Announce Type: new Abstract: LLMs often generate seemingly valid answers to flawed or ill-posed inputs. This is not due to missing knowledge: under discriminative prompting, the same models can mostly identify such issues, yet fail to refle…

  benchmarkllm
• Improving Safety Alignment via Balanced Direct Preference Optimization 重点国际前沿模型发布普通质量分 66.7 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22829v1 Announce Type: new Abstract: With the rapid development and widespread application of Large Language Models (LLMs), their potential safety risks have attracted widespread attention. Reinforcement Learning from Human Feedback (RLHF) has been…

  apibenchmarkllmtraining
• Dynamical Systems Theory Behind a Hierarchical Reasoning Model 重点国际前沿模型发布普通质量分 64.5 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22871v1 Announce Type: new Abstract: Current large language models (LLMs) primarily rely on linear sequence generation and massive parameter counts, yet they severely struggle with complex algorithmic reasoning. While recent reasoning architectures…

  benchmarkllmtraining
• Maximum Entropy Relaxation of Multi-Way Cardinality Constraints for Synthetic Population Generation 重点国际模型发布持续热议质量分 70.9 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22558v1 Announce Type: new Abstract: Generating synthetic populations from aggregate statistics is a core component of microsimulation, agent-based modeling, policy analysis, and privacy-preserving data release. Beyond classical census marginals, m…

  agentbenchmarkrelease
• Benchmarking Multi-Agent LLM Architectures for Financial Document Processing: A Comparative Study of Orchestration Patterns, Cost-Accuracy Tradeoffs and Production Scaling Strategies 重点国际模型发布普通质量分 69.9 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22651v1 Announce Type: new Abstract: The adoption of large language models (LLMs) for structured information extraction from financial documents has accelerated rapidly, yet production deployments face fundamental architectural decisions with limit…

  agentapibenchmarkllm
• Chain-of-Authorization: Internalizing Authorization into Large Language Models via Reasoning Trajectories 国际模型发布普通质量分 63.3 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22869v1 Announce Type: new Abstract: Large Language Models (LLMs) have become core cognitive components in modern artificial intelligence (AI) systems, combining internal knowledge with external context to perform complex tasks. However, LLMs typic…

  llmtraining
• Computational Arbitrage in AI Model Markets 国际模型发布持续热议质量分 70.5 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22404v1 Announce Type: new Abstract: Consider a market of competing model providers selling query access to models with varying costs and capabilities. Customers submit problem instances and are willing to pay up to a budget for a verifiable soluti…

  deepseekgptinference
• Beyond Binary Correctness: Scaling Evaluation of Long-Horizon Agents on Subjective Enterprise Tasks 国际模型发布普通质量分 68.9 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22744v1 Announce Type: new Abstract: Large language models excel on objectively verifiable tasks such as math and programming, where evaluation reduces to unit tests or a single correct answer. In contrast, real-world enterprise work is often subje…

  agentapillmreleaseworkflow
• Understanding LLM Performance Degradation in Multi-Instance Processing: The Roles of Instance Count and Context Length 国际模型发布普通质量分 67.8 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22608v1 Announce Type: new Abstract: Users often rely on Large Language Models (LLMs) for processing multiple documents or performing analysis over a number of instances. For example, analysing the overall sentiment of a number of movie reviews req…

  llm
• AgriPestDatabase-v1.0: A Structured Insect Dataset for Training Agricultural Large Language Model 国际模型发布普通质量分 67.0 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22777v1 Announce Type: new Abstract: Agricultural pest management increasingly relies on timely and accurate access to expert knowledge, yet high quality labeled data and continuous expert support remain limited, particularly for farmers operating …

  apillmqwentraining
• GitHub Trending | hsliuping/TradingAgents-CN 重点国际模型发布新增质量分 82.7 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star hsliuping / TradingAgents-CN 基于多智能体LLM的中文金融交易框架 - TradingAgents中文增强版；语言: Python；今日新增星标: 449；累计星标: 21288

  agentllm
• GitHub Trending | letta-ai/claude-subconscious 国际模型发布新增质量分 81.8 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star letta-ai / claude-subconscious Give Claude Code a subconscious；语言: TypeScript；今日新增星标: 71；累计星标: 1421

  claude
• Inside our approach to the Model Spec 国际模型发布持续热议质量分 77.5 · 前沿分 43.0 · 2026-03-25T18:00:00+08:00 · 海外前沿实验室/OpenAI News

  Learn how OpenAI’s Model Spec serves as a public framework for model behavior, balancing safety, user freedom, and accountability as AI systems advance.

  insideapproachmodelspeclearn
• X | Demis Hassabis (@demishassabis) 国际模型发布新增质量分 76.4 · 前沿分 43.0 · 2026-03-26T04:22:39+08:00 · follow-builders/X Priority Watchlist / AI Leadership

  角色: Google DeepMind CEO。 Perfect background music for flow state at 2am - made with the new Lyria 3 Pro. Google AI subscribers can try it in the @GeminiApp and developers can build with the API in @GoogleAIStudio - have fun!! Video Google DeepMind (@GoogleDee…

  apigemini
• X | Peter Yang (@petergyang) 国际模型发布持续热议质量分 76.1 · 前沿分 43.0 · 2026-03-25T12:59:16+08:00 · follow-builders/Follow Builders / X Builders Feed

  角色: I share extremely practical AI tutorials and interviews。 I hope AI agents will help us all stop obsessing about: - Specs - Designs - Issues - Roadmaps - Strategy docs - Spreadsheets and instead let us all obsess about improving the end user product and ra…

  agentapiclaudeextremelyinterviews
• X | Cat Wu (@_catwu) 国际模型发布持续热议质量分 64.5 · 前沿分 43.0 · 2026-03-25T08:02:56+08:00 · follow-builders/Follow Builders / X Builders Feed

  角色: claude code + cowork @anthropicai, prev: @dagster, @scale_ai。 Excited to see everyone at Code with Claude! What would you like to hear from us at this year's sessions? https://t.co/RxWLfs4kFo

  claude
• X | Nan Yu (@thenanyu) 国际模型发布持续热议质量分 61.7 · 前沿分 43.0 · 2026-03-25T10:29:39+08:00 · follow-builders/Follow Builders / X Builders Feed

  角色: head of product @linear。 My incentives were to get this agent thing to work so I didn’t have to do annoying work anymore! https://t.co/vI6hR0CVvq

  agentgemini
• Google launches Lyria 3 Pro music generation model 国际模型发布持续热议质量分 72.2 · 前沿分 25.0 · 2026-03-26T00:37:43+08:00 · 海外媒体/TechCrunch - Artificial Intelligence

  Google is launching Lyria 3 Pro, an upgraded music model that generates longer, more customizable tracks, as it expands AI music tools across Gemini, enterprise products, and other services.

  gemini
• The AI Hype Index: AI goes to war 国际模型发布持续热议质量分 69.5 · 前沿分 25.0 · 2026-03-25T17:00:00+08:00 · 海外媒体/MIT Technology Review - AI

  AI is at war. Anthropic and the Pentagon feuded over how to weaponize Anthropic’s AI model Claude; then OpenAI swept the Pentagon off its feet with an “opportunistic and sloppy” deal. Users quit ChatGPT in droves. People marched through London in the biggest …

  claudegpt
• Vol1.人才留言板｜第一期：这20家AI明星公司正在等你 国内模型发布新增质量分 81.7 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  图源:36kr 每周精选30+优质岗位，覆盖AI、硬科技等领域，帮你找到下一个职业爆发点。 本期合作企业：红杉中国、高瓴、鼎晖、经纬被投企业｜B轮及以上占比65%。 卷首语 我们见过太多这样的场景：一家硬科技公司融完B轮，创始人却在朋友圈悄悄挂出招聘启事；一个专注大模型的算法专家，翻遍猎头推荐却找不到真正懂技术的团队。 这不是供需错配，而是信息失真。 过去十年，36氪记录了数千家公司的成长轨迹。我们看过他们最早的BP，见证他们一轮轮融资，知道哪家公司的技术栈真正扎实，哪个创始团队对人才有足够耐心。这些认知，不该只…

  agentllm产品模型融资
• 拓竹和泡泡玛特和解，免费打印大牌IP的日子正在倒计时 国内模型发布新增质量分 78.7 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  作者 | 张子怡 编辑 | 袁斯来 在野蛮增长几年后，3D打印赛道玩家们一直没有直面的“灰犀牛”，终于冲入大门。 3月中旬，3D打印企业拓竹科技（Bambu Lab）发布声明，就旗下MakerWorld平台存在的IP侵权问题向潮玩巨头泡泡玛特致歉，双方达成和解，相关未经授权的LABUBU等模型被全面下架。 这场原定于4月对簿公堂的诉讼以“闪电和解”告终，不过，它为整个消费级3D打印行业敲响的版权警钟，才刚开始回荡。 这并非拓竹第一次面对侵权纠纷。罗小黑工作室（公司名：北京寒木春华动画技术有限公司）曾于2025年底…

  产品开源模型
• 快手：预计2026年Capex约260亿元，将继续加码AI算力与基础设施投入 国内模型发布新增质量分 70.7 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  36氪获悉，在2025年第四季度业绩电话会上，快手CFO金秉表示，预计2026年集团整体Capex投入将达到约260亿元人民币。较2025年新增的110亿投入将主要用于可灵大模型及其他基础大模型的算力支撑，也包括离线数据存储处理等常规的服务器采购支出以及数据/算力中心建设工程投入。金秉表示，2026年，尽管Capex会明显增加，公司仍以继续保持全年集团层面健康稳健的自由现金流为目标。

  模型
• 程一笑：截至26年1月可灵AI ARR已超3亿美元，预计今年收入翻倍以上增长 国内模型发布新增质量分 68.2 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  36氪获悉，在2025年第四季度业绩电话会上，快手科技创始人兼首席执行官程一笑透露，截至今年1月，可灵AI的年化收入运行率（ARR）已超过3亿美元。程一笑表示，基于当前的增长势头和商业化进展，公司对可灵AI在2026年实现收入同比翻倍以上增长保持较强信心。随着模型能力持续提升、产品形态不断完善以及专业场景适配能力增强，可灵AI商业化有望进一步提速。

  产品模型

产品更新 (1)

• 从互连工程到生态协同：TE Connectivity在机器人产业链的双重角色 国内产品更新新增质量分 75.5 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  3月20日，TE Connectivity（简称 “TE” ）在苏州举办了 “机器人核心技术赋能智能工厂创新大会” 。活动通过技术交流与工厂参观等形式，邀请机器人产业链上下游企业与生态圈伙伴围绕核心技术与落地问题展开讨论，搭建起面向产业的交流与对接平台。与会讨论折射出一个共同背景：机器人正从技术验证加速走向工程化与规模化落地。 今年，政府工作报告连续第三年提出“人工智能+”，并首次提出“打造智能经济新形态”，推动重点行业领域加速商业化和规模化应用，同时将具身智能列入未来产业布局。在政策与应用场景共同推动下，机器人…

  产品

Agent/自动化 (15)

• Session Risk Memory (SRM): Temporal Authorization for Deterministic Pre-Execution Safety Gates 重点国际Agent/自动化持续热议质量分 71.9 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22350v1 Announce Type: new Abstract: Deterministic pre-execution safety gates evaluate whether individual agent actions are compatible with their assigned roles. While effective at per-action authorization, these systems are structurally blind to d…

  agentbenchmarkinferencetraining
• Empirical Comparison of Agent Communication Protocols for Task Orchestration 国际Agent/自动化普通质量分 63.5 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22823v1 Announce Type: new Abstract: Context. Nowadays, artificial intelligence agent systems are transforming from single-tool interactions to complex multi-agent orchestrations. As a result, two competing communication protocols have emerged: a t…

  agentbenchmark
• CoMaTrack: Competitive Multi-Agent Game-Theoretic Tracking with Vision-Language-Action Models 国际Agent/自动化普通质量分 63.2 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22846v1 Announce Type: new Abstract: Embodied Visual Tracking (EVT), a core dynamic task in embodied intelligence, requires an agent to precisely follow a language-specified target. Yet most existing methods rely on single-agent imitation learning,…

  agentbenchmarktraining
• Intelligence Inertia: Physical Principles and Applications 国际Agent/自动化持续热议质量分 69.9 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22347v1 Announce Type: new Abstract: While Landauer's principle establishes the fundamental thermodynamic floor for information erasure and Fisher Information provides a metric for local curvature in parameter space, these classical frameworks func…

  agenttraining
• The Efficiency Attenuation Phenomenon: A Computational Challenge to the Language of Thought Hypothesis 国际Agent/自动化持续热议质量分 69.8 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22312v1 Announce Type: new Abstract: This paper computationally investigates whether thought requires a language-like format, as posited by the Language of Thought (LoT) hypothesis. We introduce the ``AI Private Language'' thought experiment: if tw…

  agent
• From Static Templates to Dynamic Runtime Graphs: A Survey of Workflow Optimization for LLM Agents 国际Agent/自动化持续热议质量分 69.6 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22386v1 Announce Type: new Abstract: Large language model (LLM)-based systems are becoming increasingly popular for solving tasks by constructing executable workflows that interleave LLM calls, information retrieval, tool use, code execution, memor…

  agentllmworkflow
• ABSTRAL: Automatic Design of Multi-Agent Systems Through Iterative Refinement and Topology Optimization 国际Agent/自动化普通质量分 64.6 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22791v1 Announce Type: new Abstract: How should multi-agent systems be designed, and can that design knowledge be captured in a form that is inspectable, revisable, and transferable? We introduce ABSTRAL, a framework that treats MAS architecture as…

  agentgptrelease
• Learning What Matters Now: Dynamic Preference Inference under Contextual Shifts 国际Agent/自动化普通质量分 63.0 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22813v1 Announce Type: new Abstract: Humans often juggle multiple, sometimes conflicting objectives and shift their priorities as circumstances change, rather than following a fixed objective function. In contrast, most computational decision-makin…

  agentinference
• GitHub Trending | ruvnet/ruflo 重点国际Agent/自动化新增质量分 85.4 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star ruvnet / ruflo 🌊 The leading agent orchestration platform for Claude. Deploy intelligent multi-agent swarms, coordinate autonomous workflows, and build conversational AI systems. Features enterprise-grade architecture, distributed swarm intelligence, RAG…

  agentclaudeworkflow
• GitHub Trending | bytedance/deer-flow 重点国际Agent/自动化新增质量分 83.8 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star bytedance / deer-flow An open-source SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours.；语言: Pyt…

  agent
• Introducing the OpenAI Safety Bug Bounty program 国际Agent/自动化持续热议质量分 63.5 · 前沿分 46.2 · 2026-03-25T08:00:00+08:00 · 海外前沿实验室/OpenAI News

  OpenAI launches a Safety Bug Bounty program to identify AI abuse and safety risks, including agentic vulnerabilities, prompt injection, and data exfiltration.

  agent
• Agentic commerce runs on truth and context 国际Agent/自动化持续热议质量分 70.2 · 前沿分 28.2 · 2026-03-25T19:48:13+08:00 · 海外媒体/MIT Technology Review - AI

  Imagine telling a digital agent, “Use my points and book a family trip to Italy. Keep it within budget, pick hotels we’ve liked before, and handle the details.” Instead of returning a list of links, the agent assembles an itinerary and executes the purchase. …

  agent
• Granola raises $125M, hits $1.5B valuation as it expands from meeting notetaker to enterprise AI app 国际Agent/自动化持续热议质量分 67.8 · 前沿分 25.0 · 2026-03-25T22:48:40+08:00 · 海外媒体/TechCrunch - Artificial Intelligence

  Granola's valuation jumped from $250 million to $1.5 billion with this round, and it has added more support for AI agents after users previously complained.

  agent
• AutoGenesis：基于 AI + MCP 的跨平台自动化测试实践 国内Agent/自动化新增质量分 70.6 · 前沿分 25.0 · 2026-03-26T04:43:49+08:00 · 国内聚合网站/InfoQ 中国

  点击查看原文>

  autogenesis
• 给 Agent 做“CT”：大规模 Agent 的可观测与质量保障体系｜QCon北京 国内Agent/自动化普通质量分 58.1 · 前沿分 25.0 · 2026-03-25T18:00:00+08:00 · 国内聚合网站/InfoQ 中国

  点击查看原文>

  agent

开源项目 (9)

• Memory Bear AI Memory Science Engine for Multimodal Affective Intelligence: A Technical Report 重点国际前沿开源项目持续热议质量分 72.1 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22306v1 Announce Type: new Abstract: Affective judgment in real interaction is rarely a purely local prediction problem. Emotional meaning often depends on prior trajectory, accumulated context, and multimodal evidence that may be weak, noisy, or i…

  benchmarkinference
• Can LLM Agents Generate Real-World Evidence? Evaluating Observational Studies in Medical Databases 国际开源项目普通质量分 66.5 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22767v1 Announce Type: new Abstract: Observational studies can yield clinically actionable evidence at scale, but executing them on real-world databases is open-ended and requires coherent decisions across cohort construction, analysis, and reporti…

  agentapibenchmarkllm
• STEM Agent: A Self-Adapting, Tool-Enabled, Extensible Architecture for Multi-Protocol AI Agent Systems 国际开源项目持续热议质量分 68.0 · 前沿分 60.2 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22359v1 Announce Type: new Abstract: Current AI agent frameworks commit early to a single interaction protocol, a fixed tool integration strategy, and static user models, limiting their deployment across diverse interaction paradigms. To address th…

  agent
• GitHub Trending | Crosstalk-Solutions/project-nomad 重点国际开源项目新增质量分 88.7 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star Crosstalk-Solutions / project-nomad Project N.O.M.A.D, is a self-contained, offline survival computer packed with critical tools, knowledge, and AI to keep you informed and empowered—anytime, anywhere.；语言: TypeScript；今日新增星标: 1717；累计星标: 16560

  crosstalksolutionsprojectnomadstar
• GitHub Trending | usestrix/strix 重点国际开源项目新增质量分 86.1 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star usestrix / strix Open-source AI hackers to find and fix your app’s vulnerabilities.；语言: Python；今日新增星标: 215；累计星标: 21703

  usestrixstrixstaropensource
• GitHub Trending | BerriAI/litellm 重点国际开源项目新增质量分 84.2 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Sponsor Star BerriAI / litellm Python SDK, Proxy Server (AI Gateway) to call 100+ LLM APIs in OpenAI (or native) format, with cost tracking, guardrails, loadbalancing and logging. [Bedrock, Azure, OpenAI, VertexAI, Cohere, Anthropic, Sagemaker, HuggingFace, V…

  apillm
• GitHub Trending | mvanhorn/last30days-skill 重点国际开源项目新增质量分 82.2 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star mvanhorn / last30days-skill AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary；语言: Python；今日新增星标: 1342；累计星标: 7509

  agent
• GitHub Trending | supermemoryai/supermemory 国际开源项目新增质量分 81.3 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star supermemoryai / supermemory Memory engine and app that is extremely fast, scalable. The Memory API for the AI era.；语言: TypeScript；今日新增星标: 809；累计星标: 19178

  api
• X | Zara Zhang (@zarazhangrui) 国际开源项目持续热议质量分 72.0 · 前沿分 43.0 · 2026-03-25T13:26:04+08:00 · follow-builders/Follow Builders / X Builders Feed

  角色: Builder. Dangerously skips permissions. Harvard’17. GitHub: https://t.co/KCuEajezlL YouTube: https://t.co/8xzbGWtf6w。 GitHub is the new Substack Code is a form of self-expression

  zarazhangzarazhangruibuilderdangerously

创业/融资 (4)

• X | Nikunj Kothari (@nikunj) 国际创业/融资普通质量分 58.0 · 前沿分 43.0 · 2026-03-25T14:03:33+08:00 · follow-builders/Follow Builders / X Builders Feed

  角色: partner @fpvventures - investing in seed/A. previous: early hire @meter, @opendoor, @atlassian & others. love @shimolee…。 @NotionHQ @RamP @brexHQ @harvey @EvidenceOpen @perplexity_ai @conductor_build @Railway @claudeai Obligatory - if you are building AI …

  claude
• 氪星晚报｜拼多多官宣“新拼姆”方案：开启品牌自营，一期已注资；阿里获评全球最具创新力AI企业，是唯一入选的亚洲公司；三星加入韩国政府倡导的车辆轮换限行计划，以应对能源短缺 国内创业/融资新增质量分 78.2 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  大公司： 斯里兰卡国家支付网络LankaPay宣布：40万商户已支持微信支付 36氪获悉，3月24日，斯里兰卡国家支付网络LankaPay宣布正式接入微信支付。即日起，中国游客可在斯里兰卡全岛超过40万家支持LANKAQR二维码的商户中，直接使用微信支付进行消费。目前，微信跨境支付已覆盖全球78个国家和地区，支持36种货币，覆盖餐饮、百货、商超便利、交通出行、景点购票、留学缴费等丰富场景。 拼多多官宣“新拼姆”方案：开启品牌自营，一期已注资150亿 36氪获悉，3月25日，拼多多集团正式宣布组建“新拼姆”，未来三…

  产品开源模型融资
• 生产速度提升四倍，推出桌面级金属3D打印机，「融速科技」完成近亿元A轮融资 | 36氪首发 国内创业/融资新增质量分 77.1 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  文 | 张冰冰 编辑 | 阿至 36氪获悉，苏州融速智造科技有限公司（以下简称「融速科技」） 宣布完成近亿元A轮融资，投资方为河南豫资集团、苏高新融享基金。 资金将用于进一步优化产品技术及产线建设，根据市场需求提升交付能力和生产规模。 「融速科技」成立于2020年，创始人徐方达为英国巴斯大学博士后，主要从事大尺寸增材方向研究。「融速科技」聚焦新一代DED送丝3D打印技术，面向航空航天、钢铁冶金、石油化工等工业领域中大尺寸零件制造需求，提供打印、装备、器件三大类型服务，其中STAR系列产品已获得来自中航、中车、中船…

  产品模型融资
• IDG资本领投天使轮，阿里系+清华具身项目半年融超1亿元｜36氪首发 国内创业/融资新增质量分 74.6 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  36氪获悉，工业具身智能企业光象科技，已完成种子轮、天使轮及天使+轮多轮融资，累计金额超1亿元人民币，由财务投资机构 IDG 资本、东方富海联合领投，机器人产业资本埃夫特、零一创投、达泰资本、光源 L2F创业者基金跟投。 据悉，企业融资资金将主要用于公司具身机器人的核心技术研发、产品化推进及商业化交付工作。 光象科技成立于 2025 年 4 月，由前阿里巴巴高德技术总监张涛，与清华大学教授、人工智能领域专家李升波联合创办。目前，光象科技已成为全球多家汽车主机厂的具身智能战略合作方，基于让“机器人自我学习”的具身模…

  产品模型融资

公司动态 (1)

• 从高中社团起步，三个00后天才研发网络勒索病毒“疫苗”，帮企业从黑客手中“抢”数据 | 水下项目 国内公司动态新增质量分 73.4 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  “你的文件已被加密，72小时内支付赎金，否则数据将永久销毁”——这是许多人都曾遭遇过的网络勒索（Cyberextortion）的典型场景。 网络勒索是一种常见的网络犯罪，黑客通过向企业或个人的电脑系统植入勒索病毒，使其无法正常打开或运行，从而勒索赎金。Orange Cyberdefense报告显示，网络勒索犯罪在全球呈爆炸式增长，2020年以来受害者数量增至三倍，受影响组织超过1.9万个。 作为国内最早一批专攻网络勒索防御的公司，「思而听」是目前全球少数具备勒索病毒深度应急防治能力的公司之一。思而听成立于2022…

  产品模型

行业趋势 (1)

• Reliable Classroom AI via Neuro-Symbolic Multimodal Reasoning 重点国际前沿行业趋势普通质量分 64.7 · 前沿分 79.4 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22793v1 Announce Type: new Abstract: Classroom AI is rapidly expanding from low-level perception toward higher-level judgments about engagement, confusion, collaboration, and instructional quality. Yet classrooms are among the hardest real-world se…

  apibenchmarkinference

教程/测评 (0)

• 该分类暂无数据

热门观点 (0)

• 该分类暂无数据

多模态进展 (0)

• 该分类暂无数据

编程助手 (0)

• 该分类暂无数据

视频生成/图像生成/语音生成 (0)

• 该分类暂无数据

爆款案例 (0)

• 该分类暂无数据

趋势观察 (0)

• 该分类暂无数据

工具盘点 (0)

• 该分类暂无数据

其他重要情报 (4)

• Dynamic Fusion-Aware Graph Convolutional Neural Network for Multimodal Emotion Recognition in Conversations 国际其他重要情报持续热议质量分 64.3 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22345v1 Announce Type: new Abstract: Multimodal emotion recognition in conversations (MERC) aims to identify and understand the emotions expressed by speakers during utterance interaction from multiple modalities (e.g., text, audio, images, etc.). …

  inference
• X | Garry Tan (@garrytan) 国际其他重要情报持续热议质量分 68.2 · 前沿分 43.0 · 2026-03-25T13:25:22+08:00 · follow-builders/Follow Builders / X Builders Feed

  角色: President & CEO @ycombinator —Founder https://t.co/7aoJjp1iIK—designer/engineer who helps founders—SF Dem accelerating …。 Once you say “replace developers” it’s the wrong ideaspace Cars didn’t replace walking, you still have to walk a little bit AND It le…

  7aojjp1iikdesignerfoundergarrygarrytan
• The AI skills gap is here, says AI company, and power users are pulling ahead 国际其他重要情报新增质量分 75.6 · 前沿分 25.0 · 2026-03-26T05:44:13+08:00 · 海外媒体/TechCrunch - Artificial Intelligence

  Anthropic finds AI isn’t replacing jobs yet, but early data shows growing inequality as experienced users gain an edge, raising concerns about future displacement and workforce divides.

  skillsheresayscompanypower
• 为什么 OpenAI，必须砍掉 Sora？ - 新浪网 国内其他重要情报持续热议质量分 51.6 · 前沿分 25.0 · 2026-03-25T11:51:12+08:00 · 国内平台/Google News - 视频号 AI

  为什么 OpenAI，必须砍掉 Sora？ 新浪网

  openaisora

去重合并事件列表

• X | Peter Yang (@petergyang)覆盖 3 个来源 · Follow Builders / X Builders Feed, Follow Builders / X Builders Feed, Follow Builders / X Builders Feed
• X | Garry Tan (@garrytan)覆盖 3 个来源 · Follow Builders / X Builders Feed, Follow Builders / X Builders Feed, Follow Builders / X Builders Feed
• X | Nan Yu (@thenanyu)覆盖 2 个来源 · Follow Builders / X Builders Feed, Follow Builders / X Builders Feed

值得重点关注的趋势/项目

• Memory Bear AI Memory Science Engine for Multimodal Affective Intelligence: A Technical Report 重点国际前沿开源项目持续热议质量分 72.1 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22306v1 Announce Type: new Abstract: Affective judgment in real interaction is rarely a purely local prediction problem. Emotional meaning often depends on prior trajectory, accumulated context, and multimodal evidence that may be weak, noisy, or i…

  benchmarkinference
• Reliable Classroom AI via Neuro-Symbolic Multimodal Reasoning 重点国际前沿行业趋势普通质量分 64.7 · 前沿分 79.4 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22793v1 Announce Type: new Abstract: Classroom AI is rapidly expanding from low-level perception toward higher-level judgments about engagement, confusion, collaboration, and instructional quality. Yet classrooms are among the hardest real-world se…

  apibenchmarkinference
• Bridging the Know-Act Gap via Task-Level Autoregressive Reasoning 重点国际前沿模型发布普通质量分 69.8 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22619v1 Announce Type: new Abstract: LLMs often generate seemingly valid answers to flawed or ill-posed inputs. This is not due to missing knowledge: under discriminative prompting, the same models can mostly identify such issues, yet fail to refle…

  benchmarkllm
• GitHub Trending | Crosstalk-Solutions/project-nomad 重点国际开源项目新增质量分 88.7 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star Crosstalk-Solutions / project-nomad Project N.O.M.A.D, is a self-contained, offline survival computer packed with critical tools, knowledge, and AI to keep you informed and empowered—anytime, anywhere.；语言: TypeScript；今日新增星标: 1717；累计星标: 16560

  crosstalksolutionsprojectnomadstar
• Improving Safety Alignment via Balanced Direct Preference Optimization 重点国际前沿模型发布普通质量分 66.7 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22829v1 Announce Type: new Abstract: With the rapid development and widespread application of Large Language Models (LLMs), their potential safety risks have attracted widespread attention. Reinforcement Learning from Human Feedback (RLHF) has been…

  apibenchmarkllmtraining
• Session Risk Memory (SRM): Temporal Authorization for Deterministic Pre-Execution Safety Gates 重点国际Agent/自动化持续热议质量分 71.9 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22350v1 Announce Type: new Abstract: Deterministic pre-execution safety gates evaluate whether individual agent actions are compatible with their assigned roles. While effective at per-action authorization, these systems are structurally blind to d…

  agentbenchmarkinferencetraining
• Maximum Entropy Relaxation of Multi-Way Cardinality Constraints for Synthetic Population Generation 重点国际模型发布持续热议质量分 70.9 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22558v1 Announce Type: new Abstract: Generating synthetic populations from aggregate statistics is a core component of microsimulation, agent-based modeling, policy analysis, and privacy-preserving data release. Beyond classical census marginals, m…

  agentbenchmarkrelease
• Dynamical Systems Theory Behind a Hierarchical Reasoning Model 重点国际前沿模型发布普通质量分 64.5 · 前沿分 73.0 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22871v1 Announce Type: new Abstract: Current large language models (LLMs) primarily rely on linear sequence generation and massive parameter counts, yet they severely struggle with complex algorithmic reasoning. While recent reasoning architectures…

  benchmarkllmtraining
• GitHub Trending | usestrix/strix 重点国际开源项目新增质量分 86.1 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star usestrix / strix Open-source AI hackers to find and fix your app’s vulnerabilities.；语言: Python；今日新增星标: 215；累计星标: 21703

  usestrixstrixstaropensource
• Benchmarking Multi-Agent LLM Architectures for Financial Document Processing: A Comparative Study of Orchestration Patterns, Cost-Accuracy Tradeoffs and Production Scaling Strategies 重点国际模型发布普通质量分 69.9 · 前沿分 66.6 · 2026-03-25T12:00:00+08:00 · 海外前沿论文/arXiv cs.AI

  arXiv:2603.22651v1 Announce Type: new Abstract: The adoption of large language models (LLMs) for structured information extraction from financial documents has accelerated rapidly, yet production deployments face fundamental architectural decisions with limit…

  agentapibenchmarkllm
• GitHub Trending | ruvnet/ruflo 重点国际Agent/自动化新增质量分 85.4 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Star ruvnet / ruflo 🌊 The leading agent orchestration platform for Claude. Deploy intelligent multi-agent swarms, coordinate autonomous workflows, and build conversational AI systems. Features enterprise-grade architecture, distributed swarm intelligence, RAG…

  agentclaudeworkflow
• GitHub Trending | BerriAI/litellm 重点国际开源项目新增质量分 84.2 · 前沿分 51.0 · 2026-03-26T07:03:16+08:00 · github/GitHub Trending / Daily Star Growth

  Sponsor Star BerriAI / litellm Python SDK, Proxy Server (AI Gateway) to call 100+ LLM APIs in OpenAI (or native) format, with cost tracking, guardrails, loadbalancing and logging. [Bedrock, Azure, OpenAI, VertexAI, Cohere, Anthropic, Sagemaker, HuggingFace, V…

  apillm

国内补充情报（下方）

• Vol1.人才留言板｜第一期：这20家AI明星公司正在等你 国内模型发布新增质量分 81.7 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  图源:36kr 每周精选30+优质岗位，覆盖AI、硬科技等领域，帮你找到下一个职业爆发点。 本期合作企业：红杉中国、高瓴、鼎晖、经纬被投企业｜B轮及以上占比65%。 卷首语 我们见过太多这样的场景：一家硬科技公司融完B轮，创始人却在朋友圈悄悄挂出招聘启事；一个专注大模型的算法专家，翻遍猎头推荐却找不到真正懂技术的团队。 这不是供需错配，而是信息失真。 过去十年，36氪记录了数千家公司的成长轨迹。我们看过他们最早的BP，见证他们一轮轮融资，知道哪家公司的技术栈真正扎实，哪个创始团队对人才有足够耐心。这些认知，不该只…

  agentllm产品模型融资
• 拓竹和泡泡玛特和解，免费打印大牌IP的日子正在倒计时 国内模型发布新增质量分 78.7 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  作者 | 张子怡 编辑 | 袁斯来 在野蛮增长几年后，3D打印赛道玩家们一直没有直面的“灰犀牛”，终于冲入大门。 3月中旬，3D打印企业拓竹科技（Bambu Lab）发布声明，就旗下MakerWorld平台存在的IP侵权问题向潮玩巨头泡泡玛特致歉，双方达成和解，相关未经授权的LABUBU等模型被全面下架。 这场原定于4月对簿公堂的诉讼以“闪电和解”告终，不过，它为整个消费级3D打印行业敲响的版权警钟，才刚开始回荡。 这并非拓竹第一次面对侵权纠纷。罗小黑工作室（公司名：北京寒木春华动画技术有限公司）曾于2025年底…

  产品开源模型
• 氪星晚报｜拼多多官宣“新拼姆”方案：开启品牌自营，一期已注资；阿里获评全球最具创新力AI企业，是唯一入选的亚洲公司；三星加入韩国政府倡导的车辆轮换限行计划，以应对能源短缺 国内创业/融资新增质量分 78.2 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  大公司： 斯里兰卡国家支付网络LankaPay宣布：40万商户已支持微信支付 36氪获悉，3月24日，斯里兰卡国家支付网络LankaPay宣布正式接入微信支付。即日起，中国游客可在斯里兰卡全岛超过40万家支持LANKAQR二维码的商户中，直接使用微信支付进行消费。目前，微信跨境支付已覆盖全球78个国家和地区，支持36种货币，覆盖餐饮、百货、商超便利、交通出行、景点购票、留学缴费等丰富场景。 拼多多官宣“新拼姆”方案：开启品牌自营，一期已注资150亿 36氪获悉，3月25日，拼多多集团正式宣布组建“新拼姆”，未来三…

  产品开源模型融资
• 生产速度提升四倍，推出桌面级金属3D打印机，「融速科技」完成近亿元A轮融资 | 36氪首发 国内创业/融资新增质量分 77.1 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  文 | 张冰冰 编辑 | 阿至 36氪获悉，苏州融速智造科技有限公司（以下简称「融速科技」） 宣布完成近亿元A轮融资，投资方为河南豫资集团、苏高新融享基金。 资金将用于进一步优化产品技术及产线建设，根据市场需求提升交付能力和生产规模。 「融速科技」成立于2020年，创始人徐方达为英国巴斯大学博士后，主要从事大尺寸增材方向研究。「融速科技」聚焦新一代DED送丝3D打印技术，面向航空航天、钢铁冶金、石油化工等工业领域中大尺寸零件制造需求，提供打印、装备、器件三大类型服务，其中STAR系列产品已获得来自中航、中车、中船…

  产品模型融资
• 从互连工程到生态协同：TE Connectivity在机器人产业链的双重角色 国内产品更新新增质量分 75.5 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  3月20日，TE Connectivity（简称 “TE” ）在苏州举办了 “机器人核心技术赋能智能工厂创新大会” 。活动通过技术交流与工厂参观等形式，邀请机器人产业链上下游企业与生态圈伙伴围绕核心技术与落地问题展开讨论，搭建起面向产业的交流与对接平台。与会讨论折射出一个共同背景：机器人正从技术验证加速走向工程化与规模化落地。 今年，政府工作报告连续第三年提出“人工智能+”，并首次提出“打造智能经济新形态”，推动重点行业领域加速商业化和规模化应用，同时将具身智能列入未来产业布局。在政策与应用场景共同推动下，机器人…

  产品
• IDG资本领投天使轮，阿里系+清华具身项目半年融超1亿元｜36氪首发 国内创业/融资新增质量分 74.6 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  36氪获悉，工业具身智能企业光象科技，已完成种子轮、天使轮及天使+轮多轮融资，累计金额超1亿元人民币，由财务投资机构 IDG 资本、东方富海联合领投，机器人产业资本埃夫特、零一创投、达泰资本、光源 L2F创业者基金跟投。 据悉，企业融资资金将主要用于公司具身机器人的核心技术研发、产品化推进及商业化交付工作。 光象科技成立于 2025 年 4 月，由前阿里巴巴高德技术总监张涛，与清华大学教授、人工智能领域专家李升波联合创办。目前，光象科技已成为全球多家汽车主机厂的具身智能战略合作方，基于让“机器人自我学习”的具身模…

  产品模型融资
• 从高中社团起步，三个00后天才研发网络勒索病毒“疫苗”，帮企业从黑客手中“抢”数据 | 水下项目 国内公司动态新增质量分 73.4 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  “你的文件已被加密，72小时内支付赎金，否则数据将永久销毁”——这是许多人都曾遭遇过的网络勒索（Cyberextortion）的典型场景。 网络勒索是一种常见的网络犯罪，黑客通过向企业或个人的电脑系统植入勒索病毒，使其无法正常打开或运行，从而勒索赎金。Orange Cyberdefense报告显示，网络勒索犯罪在全球呈爆炸式增长，2020年以来受害者数量增至三倍，受影响组织超过1.9万个。 作为国内最早一批专攻网络勒索防御的公司，「思而听」是目前全球少数具备勒索病毒深度应急防治能力的公司之一。思而听成立于2022…

  产品模型
• 快手：预计2026年Capex约260亿元，将继续加码AI算力与基础设施投入 国内模型发布新增质量分 70.7 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  36氪获悉，在2025年第四季度业绩电话会上，快手CFO金秉表示，预计2026年集团整体Capex投入将达到约260亿元人民币。较2025年新增的110亿投入将主要用于可灵大模型及其他基础大模型的算力支撑，也包括离线数据存储处理等常规的服务器采购支出以及数据/算力中心建设工程投入。金秉表示，2026年，尽管Capex会明显增加，公司仍以继续保持全年集团层面健康稳健的自由现金流为目标。

  模型
• AutoGenesis：基于 AI + MCP 的跨平台自动化测试实践 国内Agent/自动化新增质量分 70.6 · 前沿分 25.0 · 2026-03-26T04:43:49+08:00 · 国内聚合网站/InfoQ 中国

  点击查看原文>

  autogenesis
• 程一笑：截至26年1月可灵AI ARR已超3亿美元，预计今年收入翻倍以上增长 国内模型发布新增质量分 68.2 · 前沿分 25.0 · 2026-03-26T07:02:52+08:00 · 国内聚合网站/36氪

  36氪获悉，在2025年第四季度业绩电话会上，快手科技创始人兼首席执行官程一笑透露，截至今年1月，可灵AI的年化收入运行率（ARR）已超过3亿美元。程一笑表示，基于当前的增长势头和商业化进展，公司对可灵AI在2026年实现收入同比翻倍以上增长保持较强信心。随着模型能力持续提升、产品形态不断完善以及专业场景适配能力增强，可灵AI商业化有望进一步提速。

  产品模型
• 给 Agent 做“CT”：大规模 Agent 的可观测与质量保障体系｜QCon北京 国内Agent/自动化普通质量分 58.1 · 前沿分 25.0 · 2026-03-25T18:00:00+08:00 · 国内聚合网站/InfoQ 中国

  点击查看原文>

  agent
• 为什么 OpenAI，必须砍掉 Sora？ - 新浪网 国内其他重要情报持续热议质量分 51.6 · 前沿分 25.0 · 2026-03-25T11:51:12+08:00 · 国内平台/Google News - 视频号 AI

  为什么 OpenAI，必须砍掉 Sora？ 新浪网

  openaisora

原始来源索引

• arXiv cs.AI平台: 海外前沿论文 · 条目: 23
• GitHub Trending / Daily Star Growth平台: github · 条目: 9
• Follow Builders / X Builders Feed平台: follow-builders · 条目: 6
• TechCrunch - Artificial Intelligence平台: 海外媒体 · 条目: 3
• MIT Technology Review - AI平台: 海外媒体 · 条目: 2
• OpenAI News平台: 海外前沿实验室 · 条目: 2
• X Priority Watchlist / AI Leadership平台: follow-builders · 条目: 1
• 36氪平台: 国内聚合网站 · 条目: 9
• InfoQ 中国平台: 国内聚合网站 · 条目: 2
• Google News - 视频号 AI平台: 国内平台 · 条目: 1

抓取告警

• VentureBeat - AI · HTTPError · https://venturebeat.com/ai/feed/
• The Verge - AI · HTTPError · https://www.theverge.com/ai-artificial-intelligence/rss/index.xml