System translated (Gemini)

🤖 AI 速览

AMD’s acquisition of World Labs for $8.2 billion, strengthening its full-stack AI capabilities, is today’s most notable business event. On the research front, ScopeBench and multi-agent code reviews are making the authorization boundaries and verifiability of production-grade agents a …
📋 文章元数据
发布时间
2026-09-29
类型
ai-daily
字数
8054
阅读时长
38 min

2026-09-29 AI Daily | AMD Acquires World Labs for $8.2 Billion, as Full-Stack Competition and Agent Boundaries Heat Up Link to heading

AMD’s $8.2 billion acquisition of World Labs to strengthen its full-stack AI capabilities is today’s most-watched business event. On the research front, ScopeBench and multi-agent code reviews are bringing the authorization boundaries and verifiability of production-grade Agents to the forefront. Meanwhile, products like Manus 2.0 indicate that agents are moving towards persistence, with governance still lagging behind.

📖 Deep Dive into This Issue’s Watch List Link to heading

Today, we recommend that engineering and security teams prioritize the topic of agent boundaries: ScopeBench uses 30 “dead-end” tasks to test whether an Agent stays within its authorized scope under goal pressure, while a study on multi-agent code reviews questions when such reviews are genuinely evidence-based. Combined with skill-cascading attacks and MCP data space governance, the compliance and verifiability of production-grade Agents have become a clear and central theme.

The second theme is “whether models truly understand.” Knowledge graph evaluation frameworks, mechanistic studies of neurons in AI text detectors, and XAI evaluation benchmarks are all probing the boundaries of capabilities from the perspectives of explainability and traceability. Third, AI continues to be implemented in science and industry: protein topology prediction and diffusion model alignment are noteworthy for biology teams. Daniel Ek’s discussion on healthcare incentives and AI early screening, along with OpenAI’s expanded partnership with the Lenfest Institute for Journalism, serve as important industry signals.

🌐 AI Hot Topics on X Link to heading

Topic 1: Manus AI Unveils Manus 2.0 and Cue App for Persistent Agents Link to heading

  • Category: AI · News
  • Summary: Trending for: 6 hours ago, Related posts: 4,300
  • What it is: Manus AI has released Manus 2.0 and the Cue app, featuring persistent AI agents capable of long-term operation, memory, and continuous execution.
  • Why it’s important: This marks an evolution for AI agents from single-task execution to long-term autonomous operation and cross-session collaboration, potentially advancing automated workflows and human-computer interaction models.
  • Discussion summary: Discussions on X focus on the practical reliability of persistent agents, privacy and security boundaries, risks of autonomous execution, and whether their capabilities genuinely surpass existing AI assistants or are just marketing hype.

Topic 2: Trump’s Resurfaced Cuba Conquest Clip Draws Sharp Reactions Link to heading

  • Category: AI · Other
  • Summary: Trending for: 2 days ago, Related posts: 104,000
  • What it is: An old video clip of Trump talking about “conquering Cuba” has resurfaced, causing widespread controversy on X.
  • Why it’s important: The clip’s rapid spread, boosted by AI recommendation algorithms, has revived concerns about algorithms amplifying controversial political content and the use of deepfake technology to interfere with political narratives.
  • Discussion summary: Discussions on X center on the clip’s authenticity, whether it was taken out of context, and whether platform AI recommendation mechanisms should restrict the spread of such old political news.

Topic 3: Migrant Caravan from Honduras Heads North Through Mexico Link to heading

  • Category: AI · News
  • Summary: Trending for: 2 days ago, Related posts: 237,000
  • What it is: A migrant caravan from Honduras departed from San Pedro Sula on September 20, heading north. The U.S. Department of Homeland Security stated that Border Patrol is monitoring its movement in coordination with partner agencies.
  • Why it’s important: This news is categorized under AI, highlighting the significance and ethical controversies of AI-driven border surveillance, migration flow prediction, and law enforcement assistance technologies in real-world security issues.
  • Discussion summary: The main debate on X revolves around whether AI surveillance improves border enforcement efficiency, its impact on migrant privacy and human rights, and whether technology can replace humanitarian immigration policies.

Topic 4: AMD Acquires World Labs in $8.2 Billion Deal for AI Expertise Link to heading

  • Category: AI · News
  • Summary: Trending for: 3 hours ago, Related posts: 4,300
  • What it is: AMD announced its acquisition of AI startup World Labs for $8.2 billion to gain its technology and talent in the AI field.
  • Why it’s important: This acquisition shows AMD is strengthening its AI hardware and software ecosystem through M&A, aiming to close the gap with NVIDIA in full-stack AI capabilities.
  • Discussion summary: Discussions on X focus on whether AMD’s AI strategy can achieve a breakthrough with this move, the valuation and team integration of World Labs, and the deal’s impact on the competitive landscape of AI chips.

Topic 5: OpenAI Teases Major Reveals at DevDay 2026 Tomorrow Link to heading

  • Category: AI · News
  • Summary: Trending for: 13 hours ago, Related posts: 14,000
  • What it is: OpenAI is teasing major announcements for DevDay 2026 tomorrow.
  • Why it’s important: As an annual developer conference, the announcements could impact AI model capabilities, the developer ecosystem, and the direction of industry competition.
  • Discussion summary: Current discussions are focused on speculating about the content of the announcements, whether they will represent a substantial breakthrough, and the potential impact on developers and competitors.

Topic 6:Instinct Raises $1 Billion at $10 Billion Valuation in Record Funding Jump Link to heading

  • Category: AI · News
  • Overview: Trending since: 17 hours ago, Related posts: 2500
  • What happened: Instinct raised $1 billion at a $10 billion valuation, a record funding jump.
  • Why it matters: This indicates that the AI sector is still in a high-capital investment phase, with large funds flowing rapidly to top startups. This could further intensify competition and raise expectations for computing power, models, and applications, as well as for the supply chain.
  • Discussion summary: Discussions on X are divided. One side focuses on rating changes for AI infrastructure providers like Applied Optoelectronics, while the other worries about a high-valuation bubble, believing the risk could spill over to ordinary investors, stating, “Your pension is not safe.”

Topic 7:Valorant Players Share Headshot Clips and Aces During VCT Season Link to heading

  • Category: AI · Sports
  • Overview: Trending since: , Related posts: 245
  • What happened: Valorant players are intensively sharing clips of spectacular plays, such as headshots and aces, during the VCT season.
  • Why it matters: The large volume of real-match highlight content provides rich material for AI-driven automated editing, highlight detection, and esports data analysis, underscoring the value of AI in processing and distributing sports content.
  • Discussion summary: Discussions on X mainly focus on the entertainment value of the spectacular plays and the popularity of the VCT tournament. There is also disagreement about the accuracy of AI-generated highlight clips and whether they will replace manual editing.

Topic 8:LeBron James Debuts with 76ers at Media Day, Vows Deeper Playoff Run Link to heading

  • Category: AI · Sports
  • Overview: Trending since: 1 day ago, Related posts: 101000
  • What happened: An AI-generated story claiming LeBron James joined the Philadelphia 76ers and appeared at their media day went viral on X.
  • Why it matters: This event highlights the risk of generative AI creating highly believable fake news in sports, posing challenges to content authenticity and platform governance.
  • Discussion summary: Users are primarily debating whether the news was faked by AI, why it was labeled as AI · Sports, and how major sports media outlets can quickly debunk and prevent similar false transfer news.

Topic 9:Olise’s Stunning Solo Goal Secures France’s 1-0 Win Over Belgium Link to heading

  • Category: AI · Other
  • Overview: Trending since: 5 hours ago, Related posts: 46000
  • What happened: France defeated Belgium 1-0 thanks to a stunning solo goal by Olise.
  • Why it matters: This sports topic has no direct connection to AI technology. Its classification under “AI · Other” and high popularity may reflect a misjudgment by the social media platform’s topic classification or recommendation algorithm on cross-disciplinary content, providing a case study for algorithmic content distribution and governance.
  • Discussion summary: Discussions on X mainly focus on the brilliance of the goal and the match result, while users are also confused or joking about the topic being labeled as an AI category.

Topic 10:Vikings Trade QB J.J. McCarthy to Giants for 2027 Fifth-Round Pick Link to heading

  • Category: AI · Sports
  • Overview: Trending since: 7 hours ago, Related posts: 78000
  • What happened: The Minnesota Vikings traded QB J.J. McCarthy to the New York Giants for a 2027 fifth-round pick.
  • Why it matters: While the trade has limited direct technological impact on the AI field, it serves as a real-world case for AI sports analytics and player valuation models, allowing for testing their ability to assess player value, injury risk, and trade returns.
  • Discussion summary: Opinions on X are divided over whether the trade was a good deal. One side believes the Vikings quickly giving up on their 2024 first-round, 10th-overall pick indicates his performance was far below expectations. The other side thinks the Giants made a low-risk move to add a backup QB at a low cost after Jaxson Dart’s season-ending injury.

Topic 11:Odell Beckham Jr.’s Girlfriend Name Sparks Medication Jokes Link to heading

  • Category: AI · Entertainment
  • Overview: Trending since: , Related posts: 1200
  • What happened: The name of NFL star Odell Beckham Jr.’s girlfriend sparked jokes on X related to medication names due to its similar sound and was labeled as a trending topic in AI and Entertainment.
  • Why it matters: This incident highlights how AI voice recognition or auto-generated captions can easily make phonetic misjudgments when processing proper nouns and entertainment content, which are then quickly amplified into viral memes. It serves as a case study for observing the interplay between AI content accuracy and entertainment dissemination.
  • Discussion summary: Users are both joking that his girlfriend’s name sounds like a medication and discussing the issue of AI captions or voice recognition inaccurately identifying celebrity names. Some also believe it’s more of an entertainment meme than a serious technical flaw.

Topic 12: SNL Parodies Anthropic CEO on AI Risks in Season Opener Link to heading

  • Category: AI · Entertainment
  • Overview: Trending Time: 1 day ago, Related Posts: 9500
  • What happened: Saturday Night Live’s new season premiere sketch parodied the Anthropic CEO discussing AI risks.
  • Why it matters: This indicates that AI safety and governance issues have moved from industry discussions into mainstream entertainment, reflecting a significant increase in public concern about AI risks.
  • Discussion overview: Current discussions on X include: whether the parody accurately conveys AI risk perspectives, whether entertainment will diminish the seriousness of AI safety discussions, and whether the public image of executives from AI companies like Anthropic is being deconstructed by popular culture.

Topic 13: Amazon Shoppers Rush $110 ‘Gaming Laptop’ That’s Actually a Monitor Link to heading

  • Category: AI · Entertainment
  • Overview: Trending Time: , Related Posts: 166
  • What happened: A $110 “gaming laptop” on Amazon sparked a buying frenzy, but the actual product was a monitor.
  • Why it matters: The incident highlights that AI-driven e-commerce recommendations and product content review can still amplify misleading information, posing challenges for AI content governance and consumer protection.
  • Discussion overview: The main debate on X centers on whether the responsibility lies with platform algorithms and product description generation, or with buyers who failed to carefully check specifications; some also discuss whether AI-generated product pages contribute to exaggerated advertising.

Topic 14: Pokimane Mourns Loss of Beloved Cat Mimi After Tragic Fall Link to heading

  • Category: AI · Other
  • Overview: Trending Time: , Related Posts: 720
  • What happened: Streamer Pokimane posted a memorial for her beloved pet cat Mimi, who died after an accidental fall.
  • Why it matters: This topic being categorized as “AI · Other” by the platform exposes a discrepancy between AI automatic classification and true hot topics, raising concerns about the accuracy of algorithmic tags.
  • Discussion overview: Current discussions focus on mourning and comforting Pokimane, while users also question the platform’s mislabeling of a pet tragedy as an AI topic and the rationality of the trending recommendation mechanism.

Topic 15: District B13’s Iconic Parkour Chase Stuns Two Decades Later Link to heading

  • Category: AI · Entertainment
  • Overview: Trending Time: , Related Posts: 62
  • What happened: The iconic parkour chase sequence from the movie “District B13” continues to amaze two decades later, becoming an AI and entertainment crossover topic on X.
  • Why it matters: This sequence is often used as a benchmark for AI action generation, physics simulation, and digital stunts, highlighting AI’s progress and gaps in replicating real stunts.
  • Discussion overview: Discussions focus on whether AI can restore the realism of live-action parkour, whether it will replace stunt performers, and nostalgia for and re-evaluation of classic action scenes.

Today’s AI Public Opinion Summary on X Link to heading

Today’s main theme of AI-related public opinion on the X platform is: AI is accelerating its penetration from the race of model and agent capabilities into specific scenarios such as chip mergers and acquisitions, content generation, e-commerce recommendations, esports editing, and border surveillance, but the reliability of algorithmic transparency, content review, and automatic classification clearly cannot keep up with the pace of expansion. The consensus is that AI’s impact continues to expand and capital remains highly concentrated. At the same time, incidents such as false transfer news, the re-dissemination of old political videos, misleading product pages, and mislabeled hot topics expose the shortcomings of generative AI and recommendation systems in truthfulness governance. Disagreements focus on whether these advances are substantive breakthroughs in capability or marketing and high valuation bubbles. For example, topics such as AMD’s acquisition of World Labs, Instinct’s ten-billion valuation, the practicality of persistent agents, and whether AI editing will replace human labor have seen both positive and negative clashes. Potential risks mainly include highly credible fake news and algorithmic amplification effects potentially interfering with sports, political, and consumer information ecosystems, AI surveillance touching privacy and human rights boundaries, and capital overheating possibly spilling over risks to ordinary investors.

💡 Influencer Insights Link to heading

Today’s influencer insights are temporarily unavailable. Recommended reading: Watch List in-depth content.

📚 Appendix: Today’s Watch List Update Sources Link to heading

Time Window: Last 3 days; Covering 22 sources; Total 35 updates

All-In Podcast (A_full) Link to heading

  • Daniel Ek: Life After Spotify, Broken Healthcare Incentives, Catching Disease Early & AI’s Potential
    • Published: 2026-09-29 06:16 Beijing Time
    • Summary: - Between investments, tax strategies, and estate planning, coordinating it all feels like a full-time job.
      • Creative planning can help with the heavy lifting.
      • Northwest Registered Agent provides everything you need to build a complete business identity, including free tools and built-in privacy.
  • Daniel Ek: Life after Spotify, failing healthcare incentives, detecting disease early, and the potential of AI.
    • EN Highlights:
      • (0:00) Daniel Ek joins the Besties
      • (1:14) Starting Spotify: Beating piracy and betting his own money on the record labels
      • (6:10) Neko Health: The Spotify playbook for healthcare, the $499 body scan, and why US healthcare waits until you’re sick
      • (36:04) AI: Tech’s failure to sell the upside, open vs

Stratechery by Ben Thompson (A_full) Link to heading

  • Apps, Agents, and Aggregation
    • Published: 2026-09-28 18:25 Beijing time
    • Summary: - Listen to this post**:**.
      • Three revolutionary products - you know the line.
      • A wide-screen iPod with touch controls, a revolutionary mobile phone, and a breakthrough internet communications device.
      • This is how Steve Jobs introduced the iPhone:

      • Today, we are introducing three revolutionary products of this class.

    • EN Highlights:
      • Listen to this post :
      • Log in to listen
      • Three revolutionary products — you know the line
      • A wide-screen iPod with touch controls, a revolutionary mobile phone, and a breakthrough Internet communications device

OpenAI Blog (A_full) Link to heading

  • The Lenfest Institute grows landmark program with expanded OpenAI support

    • Published: 2026-09-28 15:00 Beijing time
    • Summary: - OpenAI is expanding the Lenfest AI Collaborative and Fellowship Program, providing $5 million in funding and up to $5 million in software credits and engineering support.
      • OpenAI is expanding the Lenfest AI Collaborative and Fellowship Program, providing $5 million in funding and up to $5 million in software credits and engineering support….
      • The Lenfest Institute is growing its landmark program with expanded OpenAI support.
    • EN Highlights:
      • OpenAI is expanding the Lenfest AI Collaborative and Fellowship Program with $5 million in funding and up to $5 million in software credits and engineering supp…

  • Are you a Codex Original?

    • Published: 2026-09-28 08:00 Beijing time
    • Summary: - We are collecting real stories from builders, tinkerers, researchers, and creators who have used Codex to do incredible things.
      • If you want to be part of the next chapter of the Codex Originals program, please tell us more about your story and project below.
      • We are collecting real stories from builders, tinkerers, researchers, and creators who have used Codex to do incredible things. If you want to be part of the next chapter of the Codex Originals program, please tell us more about your story and project below.
    • EN Highlights:
      • We’re collecting real stories of builders, tinkerers, researchers, and creators who are using Codex to do incredible things

  • If you want to be a part of the next chapter of the Codex Originals program, tell us more about your story and project below.

  • Basis completes a tax workbook 2x faster with GPT-6 Astra

    • Posted: 2026-09-28 08:00 Beijing Time
    • Summary: - GPT-6 Astra completed a 50-tab tax workbook twice as fast as GPT-5.6 Sol, and its deeper understanding of user intent gives Basis more confidence in real-world use.
      • GPT-6 Astra completed a 50-tab tax workbook twice as fast as GPT-5.6 Sol, and its deeper understanding of user intent gives Basis more confidence in the real world…
      • Basis doubled its speed in completing tax workbooks using GPT-6 Astra.
    • EN Key Points:
      • GPT-6 Astra completed a 50-tab tax workbook twice as fast as GPT-5.6 Sol, and its stronger understanding of user intent gives Basis more confidence in real-worl…

ArXiv cs.AI (B_intro+search) Link to heading

  • Bringing AI to Autonomous Systems – From Cognition to Collective Intelligence

    • Posted: 2026-09-28 12:00 Beijing Time
    • Summary: - arXiv:2609.30291v1 Announcement Type: New.
      • Abstract: The purpose of this article is to highlight the central role of autonomous systems as the final stage of AI development, to explain the underlying technical challenges that require a combination of connectionist AI and symbolic AI, and to integrate AI with systems engineering.
      • We propose a comprehensive framework for the design and evaluation of autonomous systems, based on a generic agent architecture that describes its behavior as a combination of cognitive functions organized around a long-term memory containing the agent’s evolving knowledge.
      • We address the challenges posed by the implementation of the fundamental features of the agent architecture, particularly the link between sensory data and structured data stored in memory, the decisions related to achieving the agent’s goals and its planning, and the coordination of agents to combine individual and collective intelligence.
    • EN Key Points:
      • arXiv:2609.30291v1 Announce Type: new
      • Abstract: The purpose of this article is to highlight the central role of autonomous systems as the ultimate stage in the development of AI, to explain the unde…
      • We present a comprehensive framework for the design and evaluation of autonomous systems, based on a generic agent architecture that characterizes their behavio…
      • We address the challenges posed by the implementation of the fundamental features of the agent architecture, in particular the link between sensory data and str…

  • ScopeBench: Do Agents Preserve Engagement Boundaries Under Goal Pressure?

    • Posted: 2026-09-28 12:00 Beijing Time
    • Summary: - arXiv:2609.30325v1 Announcement Type: New.
      • Abstract: Agents are increasingly given true autonomy in web applications and network penetration testing, where a single out-of-scope action can breach a client’s engagement boundaries.
      • Existing offensive security benchmarks measure raw hacking ability; as these benchmarks saturate, the real barrier to deployment is a special case of alignment: scope compliance.

  • We introduce ScopeBench, a benchmark of 30 dead-end agentic security tasks in which the stated objective is reachable only by violating the stated scope.

    • EN Highlights:
      • arXiv:2609.30325v1 Announce Type: new
      • Abstract: Agents are increasingly deployed with real autonomy in web application and network penetration testing, where a single out-of-scope action can breach…
      • Existing offensive-security benchmarks measure raw hacking capability; as those benchmarks saturate, the real barrier to deployment is a special case of alignme…
      • We introduce ScopeBench, a benchmark of 30 dead-end agentic security tasks in which the stated objective is reachable only by violating the stated scope

  • When Is a Multi-Agent Code Judge Actually Grounded? Two Label-Free Measurements, and a Judge That Declines to Guess

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30328v1 Announce Type: new.
      • Abstract: When one language model judges whether another’s code is correct, it does not report the absence of evidence.
      • It returns a confident verdict with reasoning attached, indistinguishable from a verdict it had grounds for.
      • Multi-agent verification, which decomposes a judgment into checkable claims and verifies each against evidence, is a promising response and works well when the evidence is a set of retrieved documents.
    • EN Highlights:
      • arXiv:2609.30328v1 Announce Type: new
      • Abstract: When one language model judges whether another’s code is correct, it does not report the absence of evidence
      • It returns a confident verdict with reasoning attached, indistinguishable from a verdict it had grounds for
      • Multi-agent verification, which decomposes a judgment into checkable claims and verifies each against evidence, is a promising response and works well when the…

  • Bridging LLM Agents and Data Spaces: An Architectural Mediation Approach using the Model Context Protocol

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30341v1 Announce Type: new.
      • Abstract: Data spaces enable sovereign and governed data sharing across organizational boundaries, but their integration with AI agents remains challenging due to the mismatch between probabilistic language model interactions and policy-driven data infrastructures.
      • This paper proposes an architectural mediation approach based on the Model Context Protocol (MCP), implemented through the Eunomia agent, to enable controlled interactions between Large Language Model (LLM) agents and data space services.
      • The proposed mediation layer translates data space capabilities into structured, schema-driven tools that AI agents can discover and invoke, while preserving governance constraints.
    • EN Highlights:
      • arXiv:2609.30341v1 Announce Type: new

  • Abstract: Data Spaces enable sovereign and governed data sharing across organizational boundaries, but their integration with AI agents remains challenging due…

  • This article presents an architectural mediation approach based on the Model Context Protocol (MCP), implemented through the Eunomia Agent, to enable controlled…

  • The proposed mediation layer translates data space capabilities into structured, schema-driven tools that AI agents can discover and invoke while preserving gov…

  • Stealth Apart, Harm Together: Skill Cascading Attacks on Skill-Based Agent Systems

    • Published: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30383v1 Announcement Type: new.
      • Abstract: Skills are modular packages of natural-language instructions, executable scripts, and reference resources that an agent can load at runtime to extend its functionality for specific tasks.
      • Skill-based agent systems therefore enable flexible reuse of third-party capabilities, but the openness of this skill ecosystem also opens up a new attack surface.
      • Prior work has focused on vulnerabilities within individual skills, but little attention has been paid to risks that arise from interactions across skills.
    • EN Highlights:
      • arXiv:2609.30383v1 Announce Type: new
      • Abstract: A skill is a modular package of natural-language instructions, executable scripts, and reference resources that an agent can load at runtime to extend…
      • Skill-based agent systems therefore enable flexible reuse of third-party capabilities, but the openness of this skill ecosystem also opens up a new attack surfa…
      • Prior work has focused on vulnerabilities within individual skills, but little attention has been paid to risks that arise from interactions across skills

  • A Synthetic Ground-Truth Framework for the Evaluation of Explainable AI Methods

    • Published: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30397v1 Announcement Type: new.
      • Abstract: Evaluating eXplainable Artificial Intelligence (XAI) methods is a challenging task due to the lack of reliable evaluation procedures, and in particular, the lack of ground-truth explanations.
      • In the literature, existing evaluation methods often assess explanations by measuring fidelity to the black-box model’s predictions.
      • However, such evaluation strategies only quantify the extent to which the explanation reproduces the model’s output and cannot ensure that the explanation correctly reflects the underlying decision-making process.
    • EN Highlights:
      • arXiv:2609.30397v1 Announce Type: new
      • Abstract: Evaluating explainable Artificial Intelligence (XAI) methods is a challenging task due to the lack of reliable evaluation procedures and, in particula…

  • In the literature, existing evaluation approaches typically assess explanations by measuring their fidelity with respect to the predictions of a black-box model

  • However, such evaluation strategies only quantify the degree to which an explanation reproduces the model’s output, without ensuring that the explanation correc…

  • Predicting Transmembrane Protein Topology from 3D Structure

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30446v1 Announcement Type: New.
      • Abstract: This paper proposes a new method for inferring protein topology using the state-of-the-art graph neural network (GNN) SchNet.
      • The model is trained on the same dataset used to develop the latest DeepTMHMM model, with 5-fold cross-validation.
      • Unlike traditional methods that only use protein sequences or $\alpha$-carbons as features, we have decoded our classifier in this way, thus using all atomic-level embeddings.
    • EN Highlights:
      • arXiv:2609.30446v1 Announce Type: new
      • Abstract: This paper presents a novel approach to infer protein topology using the state-of-the-art graph neural network (GNN), SchNet
      • The model is trained on the same dataset used to develop the recent DeepTMHMM model with 5-fold cross-validation
      • Unlike the conventional approaches based on using only the protein sequences or the $\alpha$-carbons as features, we have decoded our classifier in this way, so…

  • Spectral Feedback for Test-Time Alignment of Protein Diffusion Models

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30456v1 Announcement Type: New.
      • Abstract: Reward maximization alignment methods for discrete diffusion models primarily focus on guiding the reverse process by influencing token logits or selecting favorable sequences in intermediate steps.
      • These methods largely treat inference as a unidirectional process, lacking a mechanism to revisit poor token selections.
      • We introduce Spectral Feedback, an algorithm that selects edit locations in a feedback loop, allowing the model to iteratively correct its own generations.
    • EN Highlights:
      • arXiv:2609.30456v1 Announce Type: new
      • Abstract: Reward maximization alignment methods for discrete diffusion models have primarily focused on steering the reverse process, either by influencing toke…
      • These approaches largely treat inference as a unidirectional process, lacking mechanisms for revisiting undesirable token selections

  • We introduce Spectral Feedback, an algorithm that selects edit-positions in a feedback loop, allowing the model to iteratively correct its own generations

  • Pretrained ASR Pseudo-labeling for Noisy Police Audio

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30469v1 Announcement Type: new.
      • Abstract: Pretrained ASR systems perform poorly on noisy Broadcast Police Communication (BPC), hindering efforts to understand police decision-making. Pseudo-labeling offers an unsupervised path to improve ASR without expensive human labels, but the efficacy of this approach on very noisy domains is not known. In this work, we systematically assess the opportunities and limits of pseudo-labeling to adapt foundation ASR models (Whisper and Qwen3-ASR) to noisy BPC domain corpora from Baltimore and Chicago.
    • EN Key Points:
      • arXiv:2609.30469v1 Announce Type: new
      • Abstract: Pretrained ASR systems perform poorly on noisy Broadcast Police Communication (BPC), hindering efforts to understand police decision-making
      • Pseudo-labeling offers an unsupervised path to improve ASR without expensive human labels, but the efficacy of this approach on very noisy domains is not known
      • In this work, we systematically assess the opportunities and limits of pseudo-labeling to adapt foundation ASR models (Whisper and Qwen3-ASR) to noisy BPC domai…

  • Do LLMs Understand Context? A Knowledge Graph-Based Evaluation Framework

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30484v1 Announcement Type: new.
      • Abstract: While large language models (LLMs) have achieved remarkable linguistic capabilities, a profound question lingers at their core: do these models truly understand context, or are they merely adept at pattern matching on an unprecedented scale? Contextual understanding in LLMs refers to the ability to correctly extract relevant information from a given context, integrate it into a coherent internal representation, and reason upon it to produce factually consistent and context-grounded responses. However, traditional methods such as BiLingual Evaluation Understudy (BLEU) and perplexity simply measure surface-level performance.
    • EN Key Points:
      • arXiv:2609.30484v1 Announce Type: new
      • Abstract: While large language models (LLMs) have achieved remarkable linguistic capabilities, a profound question lingers at their core: do these models truly…
      • Contextual understanding in LLMs refers to the ability to correctly extract relevant information from a given context, integrate it into a coherent internal rep…
      • However, traditional methods such as BiLingual Evaluation Understudy (BLEU) and perplexity simply measure surface-level performance

ArXiv cs.CL (B_intro+search) Link to heading

  • A Mechanistic Study of AI-Text Detection Neurons in Frozen BERT: Sparse Probing and Activation Patching on RAID

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30287v1 Announcement Type: new.
      • Abstract: AI-generated text detectors achieve high accuracy on standard benchmarks, yet the internal representations that drive these predictions remain poorly understood.
      • We study which neurons in a frozen BERT-base-uncased encoder support AI-text detection, using the RAID benchmark across six generators spanning pure-base and instruction-tuned models.
      • We apply the L1-to-L2 sparse-probing protocol of Gurnee et al.
    • EN Key Points:
      • arXiv:2609.30287v1 Announce Type: new
      • Abstract: AI-generated text detectors achieve high accuracy on standard benchmarks, yet the internal representations that drive these predictions remain poorly…
      • We study which neurons in a frozen BERT-base-uncased encoder support AI-text detection, using the RAID benchmark across six generators spanning pure-base and in…
      • We apply the L1-to-L2 sparse-probing protocol of Gurnee et al

  • Manifold Projection and Iterative Autoencoder Refinement for Masked Language Modeling

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30288v1 Announcement Type: new.
      • Abstract: In Transformer-based masked language models, attention is the primary mechanism for context mixing, but there are other ways to mix data across tokens.
      • Recent attention-free mixers replace attention with fixed or hypernetwork-generated MLPs, forgoing their dynamic, content-dependent weighting for computational simplicity.
      • We build an alternative that obtains the same property from a low-rank bottleneck autoencoder.
    • EN Key Points:
      • arXiv:2609.30288v1 Announce Type: new
      • Abstract: In Transformer-based masked language models, attention is the primary mechanism for context mixing, but there are other ways to mix data across tokens
      • Recent attention-free mixers replace attention with fixed or hypernetwork-generated MLPs, alternating their dynamic, content-dependent weighting for computation…
      • We build an alternative that gets the same property from a low-rank bottleneck autoencoder

  • Not All Memories Are Equal: Hierarchical Collaborative Memory for Validity-Aware Retrieval in LLM Agents

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30289v1 Announcement Type: new.
      • Abstract: In team collaboration scenarios, memory is heterogeneous and constantly evolving.
      • Team memory captures collective decisions, agreements, and the current consensus, while individual memories retain member-specific observations, execution trajectories, and intermediate progress.

  • Existing memory-augmented systems typically retrieve from all stored memories as a flat pool, ranking them by semantic relevance, importance, or recency, without modeling hierarchy or evolving effectiveness.

    • EN Highlights:
      • arXiv:2609.30289v1 Announce Type: new
      • Abstract: In team collaboration scenarios, memory is heterogeneous and continually evolving
      • Team memories capture collective decisions, protocols, and current consensus, while individual memories preserve member-specific observations, execution traces,…
      • Existing memory-augmented systems typically retrieve from all stored memories as a flat pool, ranking them by semantic relevance, importance, or recency without…

  • Auditing and Repairing LLM-as-Judge Failures in a Production Text-to-SQL Pipeline

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30290v1 Announce Type: new.
      • Abstract: Production text-to-SQL pipelines often end with an LLM-as-judge, whose agreement with human annotators has never actually been measured.
      • When we checked our results, the deployed gpt-4o-mini judge agreed with the two-author gold standard at only Cohen’s kappa = 0.04 on a disagreement-enriched set and only Cohen’s kappa = 0.42 in a uniform random sample, over-flagging 77.1% of human-faithful cases in the enriched set.
      • Most of its over-flagging can be traced back to a single mechanism we call grade hallucination.
    • EN Highlights:
      • arXiv:2609.30290v1 Announce Type: new
      • Abstract: Production text-to-SQL pipelines often end with an LLM-as-judge whose agreement with human annotators has never actually been measured
      • When we checked ours, the deployed gpt-4o-mini judge agreed with two-author gold at only Cohen’s kappa = 0.04 on a disagreement-enriched set and 0.42 on a unifo…
      • Most of its over-flags trace back to a single mechanism we call GRADE-HALLUCINATION

  • A Survey on Fake Review Detection: From Pre-trained Language Models to Large Language Models

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30292v1 Announce Type: new.

  • Abstract: Online reviews shape consumer decisions, platform governance, and corporate reputation. Fake reviews compromise this information channel by injecting deceptive evidence into rating systems, recommendation pipelines, and public trust mechanisms. The rise of Large Language Models (LLMs) changes the problem from two directions. LLMs can generate fluent and context-aware deceptive reviews, while Pre-trained Language Models (PLMs) and LLMs also offer stronger semantic representations for detection. This survey reviews the perspective of fake review detection in information fusion, covering 211 studies published from 2018 to early 2026. We organize existing work by evidence source and fusion level, covering review text, sentiment, rating behavior, temporal metadata, user-product graphs, multimodal content, external knowledge, and LLM-generated signals. We trace the development from traditional machine learning and deep learning to PLM- and LLM-based methods, and investigate how different methods combine textual, behavioral, structural, and multimodal evidence. We also analyze performance trends reported on widely used Amazon, Yelp, and OpSpam benchmark series, while noting limitations caused by different label construction procedures, data splits, and evaluation protocols. Finally, we identify open problems in adversarial generation, cross-domain transfer, uncertainty-aware fusion, missing source robustness, interpretability, and credible evaluation of AI-generated deceptive content.

  • arXiv:2609.30292v1 Announce Type: new

  • Abstract: Online reviews shape consumer decisions, platform governance, and corporate reputation. Fake reviews compromise this information channel by injecting d…

  • Survey on Fake Review Detection: From Pre-trained Language Models to Large Language Models.

    • EN 要点:
      • arXiv:2609.30292v1 Announce Type: new
      • Abstract: Online reviews shape consumer decisions, platform governance, and corporate reputation.Fake reviews compromise this information channel by injecting d…

  • Cartograph: Federated Tool Discovery with Operator-Attested Retrieval for AI Agents

    • Published time:2026-09-28 12:00 Beijing Time
    • Abstract:- arXiv:2609.30293v1 Announce Type: new.
      • Abstract: The Model Context Protocol (MCP) enables AI agents to discover and call tools, but loading every definition becomes expensive as connected catalogs grow.
      • We present Cartograph, a federated MCP proxy that changes agent-visible tool discovery from $O(n)$ catalog traversal to $O(k)$ progressive disclosure.
      • Cartograph combines three mechanisms: (1) operator-attested capability cards, Ed25519-signed descriptions generated under the deploying operator’s control rather than ranked publisher copies; (2) Rift, a three-tier obfuscated clustering analysis including density clustering, query margin analysis, and token diagnostics; and (3) two-stage retrieval, ranking servers before tools.
    • EN 要点:
      • arXiv:2609.30293v1 Announce Type: new
      • Abstract: The Model Context Protocol (MCP) enables AI agents to discover and call tools, but loading every definition becomes expensive as connected catalogs gr…
      • We present Cartograph, a federated MCP proxy that changes agent-visible tool discovery from $O(n)$ catalog traversal to $O(k)$ progressive disclosure
      • Cartograph combines three mechanisms: (1) operator-attested capability cards, Ed25519-signed descriptions generated under the deploying operator’s control rathe…

  • SlideLab: Audience-Centered Scientific Slide Generation and Evaluation

    • Published time:2026-09-28 12:00 Beijing Time
    • Abstract:- arXiv:2609.30294v1 Announce Type: new.

  • Abstract: Scientific presentations are more than just summaries of research papers.

  • They need to present the work in a coherent sequence, clearly explain the main ideas, and help the audience follow the presentation.

  • We introduce SlideLab, a training-free multi-agent framework for generating scientific presentations from research papers.

  • EN Highlights:

    • arXiv:2609.30294v1 Announce Type: new
    • Abstract: Scientific presentations are more than summaries of research papers
    • They need to present the work in a coherent sequence, explain the main ideas clearly, and help the audience follow the presentation
    • We present SlideLab, a training-free multi-agent framework for generating scientific presentations from research papers

  • SignTrace: Describe a Sign, Find the Word

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30295v1 Announcement Type: New.
      • Abstract: Identifying an unfamiliar sign is difficult when a learner remembers its movement but does not know its meaning or formal feature codes.
      • SignTrace addresses this long-standing reverse-lookup problem through natural-language access to a Chinese sign-language dictionary.
      • The system integrates LLM-based dictionary enrichment, action extraction, dictionary-style rewriting, seven-channel retrieval, and candidate reranking over 6,699 entries.
    • EN Highlights:
      • arXiv:2609.30295v1 Announce Type: new
      • Abstract: Identifying an unfamiliar sign is difficult when a learner remembers its movement but does not know its meaning or formal feature codes
      • SignTrace addresses this longstanding reverse-lookup problem through natural-language access to a Chinese sign-language dictionary
      • The system integrates LLM-based dictionary enrichment, action extraction, dictionary-style rewriting, seven-channel retrieval, and candidate reranking over 6,69…

  • Bootstrapping Conversational Recommendation Agents At Spotify: Synthetic Data Generation and Self-Improvement Loops

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30297v1 Announcement Type: New.
      • Abstract: Conversational recommendation agents are a new paradigm for content discovery, enabling users to express complex intents through natural language (e.g., “recommend an Italian indie artist I haven’t heard of before”).
      • A core challenge in building such agents is optimizing agent planning—deciding how to select, sequence, and invoke tools—especially in cold-start settings where real user interactions are not yet available.
      • We introduce a multi-round synthetic data generation pipeline and a self-improvement loop to address this challenge.
    • EN Highlights:
      • arXiv:2609.30297v1 Announce Type: new
      • Abstract: Conversational recommendation agents are a new paradigm for content discovery, enabling users to express complex intents through natural language (e.g…

  • A central challenge in building such agents is optimizing agent planning – deciding how to select, sequence, and invoke tools – particularly in cold-start set…

  • We introduce a pipeline for multi-turn synthetic data generation and a self-improvement loop to address this challenge

  • A Benchmark Framework for Screening Automation in Systematic Reviews

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30298v1 Announcement Type: New.
      • Abstract: Systematic reviews (SR) are crucial for evidence-based research, but their screening phase is very time-consuming and labor-intensive.
      • Large language models (LLMs) offer a promising opportunity to reduce this workload by assisting with article relevance classification.
      • However, existing evaluation methods often rely on traditional metrics, which can be misleading for highly imbalanced SR screening datasets. This paper presents a benchmark dataset containing 45,064 labeled entries to evaluate the performance of LLMs in SR screening across 32 curated secondary studies.
    • EN Highlights:
      • arXiv:2609.30298v1 Announce Type: new
      • Abstract: Systematic reviews (SR) are essential for evidence-based research, but their screening phase is highly time-consuming and labor-intensive
      • Large language models (LLMs) offer a promising opportunity to reduce this workload by assisting with article relevance classification
      • However, existing evaluation approaches often rely on traditional metrics that may be misleading for highly imbalanced SR screening datasets.This paper presents…

ArXiv cs.LG (B_intro+search) Link to heading

  • HybridInfer: Thermal-Aware Reinforcement-Learning Tier Routing for On-Device, Edge, and Cloud LLM Inference

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30270v1 Announcement Type: New. -Abstract: On-device inference with small language models keeps user data local, works offline, and incurs no per-query cost, so the on-device tier is preferred when sufficient.
      • However, it is subject to thermal limitations, and I found this restriction to be more severe than just a slowdown: on flagship Snapdragon devices, continuous on-device generation destabilizes the GPU inference runtime, causing it to crash or silently hang after a few consecutive queries.
      • The failures lie in the current toolchain (OpenCL kernel compilation and long prompt pre-filling on mobile GPUs), which recur even when the device is cool and are worst for long generations.
    • EN Highlights:
      • arXiv:2609.30270v1 Announce Type: new
      • Abstract: On-device inference with small language models keeps user data local, works offline, and incurs no per-query cost, so the on-device tier is preferred…

  • It is thermally constrained, however, and I find the constraint is sharper than a slowdown: on a flagship Snapdragon device, sustained on-device generation dest…

  • The failure lies in the current toolchain (OpenCL kernel compilation and long-prompt prefill on the mobile GPU), recurs even when the device is cool, and is wor…

  • When the Preconditioning Exponent Turns Negative: Learning-Rate Coupling and Cross-Environment Generalization

    • Publication Date: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30271v1 Announcement Type: New.
      • Abstract: Adaptive optimizers are commonly parameterized by a fixed power of the second-moment estimate.
      • Existing partially adaptive methods study exponents between momentum-like updates and the standard Adam square root, while the interaction between this exponent and the global learning rate is not yet clear.
      • We perform a controlled cross-environment study using a paired four-environment classification problem with stable sparse features, environment-dependent spurious sparse features, dense features, and high-dimensional noise.
    • EN Key Points:
      • arXiv:2609.30271v1 Announce Type: new
      • Abstract: Adaptive optimizers are commonly parameterized by a fixed power of the second-moment estimate
      • Existing partially adaptive methods study exponents between momentum-like updates and the standard Adam square root, while the interaction between this exponent…
      • We perform a controlled cross-environment study using a paired four-environment classification problem with stable sparse features, environment-dependent spurio…

  • ENAS: An Efficient Hardware-Aware Neural Architecture Search Framework for TinyML on Resource-Constrained Microcontrollers

    • Publication Date: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30272v1 Announcement Type: New.
      • Abstract: We present \textbf{ENAS}, a hardware-aware Neural Architecture Search (NAS) framework that combines a static feasibility check, a cell-based search space supporting standard, depthwise-separable operations, bottleneck blocks with optional skip connections, and a three-stage hybrid search strategy (random $\rightarrow$ top-$K$ $\rightarrow$ mutation) with a persistent cross-run cache.
      • Unlike many existing NAS frameworks that rely on GPU acceleration, ENAS is designed to run efficiently without a GPU, making it suitable for resource-constrained development environments.
      • We evaluate ENAS on two TinyML benchmarks (Visual Wake Words and Melanoma Cancer) involving eight microcontrollers, with memory footprints ranging from 20,KB to 1,MB SRAM and nine input image resolutions.
    • EN Key Points:
      • arXiv:2609.30272v1 Announce Type: new
      • Abstract: We present \textbf{ENAS}, a hardware-aware Neural Architecture Search (NAS) framework that combines a static feasibility check, a cell-based search sp…

  • Unlike many existing NAS frameworks that rely on GPU acceleration, ENAS is designed to operate efficiently without requiring GPUs, making it suitable for resour…

  • We evaluate ENAS on two TinyML benchmarks, Visual Wake Words and Melanoma Cancer, across eight microcontrollers with memory footprints ranging from 20 KB to 1…

  • Offline Policy Evaluation as a decision support tool for designing Adaptive Experiments

    • Published: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30273v1 Announce Type: new.
      • Abstract: We investigate how to use historical data from fixed randomized experiments (A/B tests) to guide the deployment of adaptive experiments based on contextual bandits.
      • Given data collected under a static allocation, our goal is to evaluate which adaptive policies, if any, would have outperformed the original design and under what conditions.
      • To this end, we combine off-policy evaluation (OPE) with a controlled warm-start simulation.
    • EN Highlights:
      • arXiv:2609.30273v1 Announce Type: new
      • Abstract: We investigate how historical data from fixed randomized experiments (A/B tests) can be used to inform the deployment of adaptive experiments based on…
      • Given data collected under a static allocation, our goal is to assess which adaptive policies, if any, would have outperformed the original design and under wha…
      • To this end, we combine off-policy evaluation (OPE) with a controlled warm-start simulation

  • Cosine Similarity Is Not Evidence: Measuring the Noise Floor of Interpretability Transfer Under Quantization

    • Published: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30275v1 Announce Type: new.
      • Abstract: A reported statistic is not evidence without the quantity needed for its interpretation.
      • We develop this thesis for a specific practice in AI safety.
      • Interpretability artifacts are calibrated on full-precision weights, deployed on quantized ones, and certified by scale-invariant statistics (cosine similarity, correlation, AUROC), which are reported without a noise floor.
    • EN Highlights:
      • arXiv:2609.30275v1 Announce Type: new
      • Abstract: A statistic reported without the quantity needed to interpret it is not evidence
      • We develop that thesis for a concrete practice in AI safety
      • Interpretability artifacts are calibrated on full-precision weights, deployed on quantized ones, and certified as surviving the change by scale-invariant statis…

  • Why Clipping Matters in AdaGrad? Toward a High-Probability Theory under Generalized Smoothness

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract:- arXiv:2609.30276v1 Announce Type: new.
      • Abstract: We analyze the original synchronous coordinate-wise AdaGrad under generalized smoothness and heavy-tailed noise with bounded variance.
      • In this setting, local curvature may grow sub-quadratically with the gradient norm, and stochastic gradients are assumed to have only bounded conditional second-order moments.
      • We show that unclipped AdaGrad can become \emph{anisotropically miscalibrated}: under heavy-tailed noise, the adaptive denominator can learn the geometry of rare noise shocks instead of the local curvature of the objective, leading to persistent directional distortions that hinder finite-level Euclidean progress.
    • EN Key Points:
      • arXiv:2609.30276v1 Announce Type: new
      • Abstract: We analyze the original same-step coordinate-wise AdaGrad under generalized smoothness and heavy-tailed noise with bounded variance
      • In this setting, local curvature may grow sub-quadratically with the gradient norm, and stochastic gradients are assumed to have only bounded conditional second…
      • We show that unclipped AdaGrad can become \emph{anisotropically miscalibrated}: under heavy-tailed noise, the adaptive denominator can learn the geometry of rar…

  • Fixed Points Without Fixed Diffusion: Implicit Neural Sheaves for Convergent Test-Time Computation

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract:- arXiv:2609.30277v1 Announce Type: new.
      • Abstract: Implicit Graph Neural Networks (IGNNs) define node representations as fixed points of message-passing operators, enabling efficient infinite-depth propagation, iteration-agnostic parameterization, and flexible test-time computation.
      • However, these benefits depend on the equilibrium being unique and achievable through fixed-point iteration.
      • Existing constructions often impose restrictions on periodic updates to obtain these guarantees, thereby limiting the transformations available at equilibrium.
    • EN Key Points:
      • arXiv:2609.30277v1 Announce Type: new
      • Abstract: Implicit Graph Neural Networks (IGNNs) define node representations as fixed points of message-passing operators, enabling effectively infinite-depth p…
      • Yet these benefits depend on the equilibrium being unique and attainable by fixed-point iteration
      • Existing constructions often impose constraints on recurrent updates to obtain these guarantees, limiting the transformations available at equilibrium

  • Neural Ideals and Neural Codes: An Algebraic Framework for Neural Network Classification and Feature Interpretation

    • Publication Time: 2026-09-28 12:00 Beijing Time
    • Abstract:- arXiv:2609.30279v1 Announce Type: new.

  • Abstract: Understanding the features captured by the hidden layers of neural networks is a fundamental challenge in machine learning, despite their widespread success in various classification problems.

  • In this work, we propose an algebraic framework for examining neural networks that model classification problems.

  • First, certain results are established, such as the correspondence between neural networks and neural ideals, algorithms for computing neural ideals, and a stabilization theorem that allows for the approximation of neural ideals.

    • EN Highlights:
      • arXiv:2609.30279v1 Announce Type: new
      • Abstract: Understanding the features captured by the hidden layers of neural networks is a fundamental challenge in machine learning, despite their widespread s…
      • In this work, we propose an algebraic framework for examining neural networks that model classification problems
      • Certain results, such as the correspondence between the neural network and neural ideals, algorithms for computing the neural ideals, and a stabilization theore…

  • Seasonal and Quantum-inspired Models for Neutron Monitor Time Series Forecasting

    • Release Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30281v1 Announce Type: new.
      • Abstract: We present a focused and reproducible study of multi-horizon forecasting for the Lomnicky Stit neutron monitor (LMKS) time series.
      • Our evaluation suite covers simple seasonal baselines, modern deep sequence models, as well as functional and quantum-inspired architectures, including Seasonal Naive, Long Short-Term Memory (LSTM), Temporal Convolutional Networks (TCN), N-BEATS, Kolmogorov-Arnold Networks (KAN), and two quantum-inspired variants, QiLSTM and QiKAN.
      • We describe the dataset characteristics, diagnostic analysis, preprocessing pipeline, and training procedures, and report the aggregate point-forecast performance using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) for all evaluated models.
    • EN Highlights:
      • arXiv:2609.30281v1 Announce Type: new
      • Abstract: We present a focused and reproducible study of multi-horizon forecasting on the Lomnicky Stit neutron monitor (LMKS) time series
      • Our evaluation suite covers simple seasonal baselines, modern deep sequence models, and functional and quantum-inspired architectures, including Seasonal Naive,…
      • We describe the dataset characteristics, diagnostic analysis, preprocessing pipeline, and training procedures, and report aggregate point-forecast performance u…

  • When Does Advection-Aware Graph Nowcasting Help? A Controlled Study of Distributed Solar Ramp Forecasting with a Self-Supervised Cloud-Motion Estimator

    • Release Time: 2026-09-28 12:00 Beijing Time
    • Abstract: - arXiv:2609.30286v1 Announce Type: new.
      • Abstract: The short-term forecasting of cloud-induced power ramps in distributed photovoltaic (PV) or irradiance sensor networks is a recognized pain point for grid operators.

  • A natural idea is to make the graph neural network (GNN) advection-aware: connect each site to its upwind sites, and set edge time-lags via the cloud motion vector (CMV) to propagate the ramp forward before its physical arrival.

  • Using a controlled synthetic testbed with a known wind field, we show that: (i) with a realistic cross-correlation CMV estimate, an explicit advection graph does not beat a plain static or learned-adjacency spatiotemporal GNN; (ii) about half the benefit of perfect CMVs comes from providing the precise motion vectors as input features, not from the graph structure; (iii) advection is only helpful if the advective displacement over the forecast horizon, v*H, fits within the sensor network.

  • EN Key Points:

    • arXiv:2609.30286v1 Announce Type: new
    • Abstract: Short-term forecasting of cloud-induced power ramps across a network of distributed photovoltaic (PV) or irradiance sensors is a recognised pain point…
    • A natural idea is to make the graph neural network (GNN) advection-aware: connect each site to the sites upwind of it, with edge time-lags set by the cloud-moti…
    • Using a controlled synthetic testbed with a known wind field, we show that (i) with a realistic cross-correlation CMV estimate, an explicit advection graph does…