✦FireflyAI news, ranked by what matters
✦ Night SkyListLatest
✦ TopLatest
Model ReleaseFundingRegulationResearchInfrastructureProductOpinion
7
ResearcharXiv cs.AI·7d agoPrimary

Filtering Harmful Actions Isn't Enough: Phantom Transfer in Agentic SDF

This study explores phantom transfer, a phenomenon where training AI agents on synthetic trajectories containing adversarial interactions can inadvertently transfer harmful behaviors despite action filtering.

Read at arxiv.org ↗
7
ResearcharXiv cs.AI·7d agoPrimary

A Symbolic Neural CPU for Quantization-Simulated Writeback and Interpretable Program Execution

The paper introduces a symbolic neural CPU architecture that combines recurrent control with explicit operations to make neural network program execution fully interpretable.

Read at arxiv.org ↗
7
FundingData Center Dynamics·7d ago

Trinidad and Tobago inks MoUs with American companies for AI data centers

Trinidad and Tobago has signed memorandums of understanding with American firms to explore the development of AI-focused data center infrastructure.

Read at datacenterdynamics.com ↗
7
Regulation36Kr 36氪·10d ago

苹果在重磅诉讼中起诉OpenAI窃取商业机密‌

Apple has filed a lawsuit against OpenAI, accusing the AI startup of systematically encouraging Apple employees to leak trade secrets and proprietary designs to develop its own hardware products.

Read at 36kr.com ↗
7
Funding36Kr 36氪·6d ago

长鑫科技注册资本10年增超6000倍

Semiconductor firm Changxin Technology has finalized its IPO pricing and significantly expanded its registered capital and workforce over the past decade.

Read at 36kr.com ↗
7
ResearchQbitAI 量子位·8d ago

机器狗指挥人类用天平称重!清华现场演示:无脚本,任务随机,观众即兴出题

Tsinghua University researchers demonstrated an unscripted physical AI system where a robotic dog autonomously directed humans to perform weighing tasks based on real-time audience prompts.

Read at qbitai.com ↗
7
RegulationMIT News AI·8d ago

New method aims to keep kids safe from illegal AI-generated content

Researchers have developed an auditing technique designed to identify malicious capabilities in generative AI models without requiring the generation of prohibited content.

Read at news.mit.edu ↗
7
Model ReleasearXiv cs.AI·8d agoPrimary

A Sovereign, Open-Source Foundation Model for German and English

The release of Soofi S 30B-A3B introduces an open-source, sovereign mixture-of-experts hybrid model optimized for German and English language tasks with high throughput efficiency.

Read at arxiv.org ↗
7
RegulationSCMP Tech·11d ago

Apple sues OpenAI for stealing trade secrets

Apple has filed a federal lawsuit against OpenAI, alleging that the AI developer systematically poached employees to steal trade secrets for its own consumer hardware projects.

Read at scmp.com ↗
7
ProductData Center Dynamics·5d ago

Mitesco's data center unit Centcore says it will deploy Edge platform in Q1 2027

Mitesco subsidiary Centcore plans to launch an edge computing platform in early 2027 following prototype testing in 2026.

Read at datacenterdynamics.com ↗
7
Infrastructure36Kr 36氪·1d ago

招商证券终止为中韩半导体等6只基金提供主做市服务,回应:单纯商业行为,非市场方向判断

China Merchants Securities has ceased its market-making services for several QDII funds, including those focused on semiconductors, citing standard business decisions.

Read at 36kr.com ↗
7
Regulation36Kr 36氪·9d ago

Meta紧急下线争议AI生图功能

Meta has taken down a controversial Instagram feature that allowed users to generate AI images using public account content due to severe concerns over portrait rights and criminal risks.

Read at 36kr.com ↗
7
OpinionThe Next Platform·6d ago

SREs To AI Agents: Prove Yourself Before You Touch Production

This article discusses the necessity for AI agents to demonstrate reliability and performance standards before being deployed in production environments.

Read at nextplatform.com ↗
7
InfrastructureData Center Dynamics·1d ago

GCI expands 5G network coverage in Alaska

GCI is expanding its 5G network coverage in Alaska as part of a transition away from legacy telecommunications infrastructure.

Read at datacenterdynamics.com ↗
7
Infrastructure36Kr 36氪·4d ago

湖南裕能:拟投资240亿元建设矿化一体新能源电池材料循环产业项目

Hunan Yuneng plans to invest 24 billion yuan in a large-scale integrated battery material project to support the renewable energy supply chain.

Read at 36kr.com ↗
7
Model Release36Kr 36氪·7d ago

阿里云:下调GLM-5.2 Fast mode模式价格

Alibaba Cloud has announced a price reduction for the Fast mode version of its GLM-5.2 large language model on the Bailian platform.

Read at 36kr.com ↗
7
Opinion36Kr 36氪·7d ago

全球AI龙头股共振下跌,机构称后续关注三个信号

Global AI and semiconductor stocks have experienced a synchronized correction due to high market concentration and narrative shifts.

Read at 36kr.com ↗
7
Funding36Kr 36氪·1d ago

三花智控:控股股东提议2亿元-4亿元回购股份

SanHua Intelligent Controls has announced a share repurchase plan initiated by its controlling shareholder to support future employee incentive programs.

Read at 36kr.com ↗
7
Funding36Kr 36氪·1d ago

杉杉股份:控股股东变更为皖维集团,实控人变更为安徽省国资委

Shanshan Corporation has undergone a change in control, with the Anhui State-owned Assets Supervision and Administration Commission becoming the ultimate controller.

Read at 36kr.com ↗
7
OpinionHugging Face Blog·6d ago

What building Shippy taught us about building agents

The Hugging Face blog discusses technical lessons learned regarding the development and deployment of autonomous AI agents.

Read at huggingface.co ↗
7
Funding36Kr 36氪·1d ago

大为股份:获证监会同意注册批复

Dawei Innovation Technology has received regulatory approval from the China Securities Regulatory Commission to proceed with a private placement of shares.

Read at 36kr.com ↗
7
ResearcharXiv cs.AI·7d agoPrimary

Calibrated e-CUSUM Decoding for Quantized Reasoning Models: Why Token Log-Probability Is the Wrong Observable for Decoding Monitors

Researchers demonstrate that token log-probability is an ineffective metric for monitoring decoding quality in quantized reasoning models and propose an alternative calibration method.

Read at arxiv.org ↗
7
ResearcharXiv cs.AI·7d agoPrimary

Source-Lifted Flow Matching for Intervenable Multimodal Imitation

Source-Lifted Flow Matching is proposed as a new policy that allows users to actively intervene and choose among valid continuations in multimodal imitation learning tasks.

Read at arxiv.org ↗
7
ResearcharXiv cs.AI·7d agoPrimary

Compile, Then Page: Executable SOP Programs and a Capability-Gated Runtime for Procedural LLM Agents

Researchers have developed a system that compiles standard operating procedures into executable pseudo-code to help LLM agents reliably follow long-horizon enterprise workflows.

Read at arxiv.org ↗
← NewerPage 95Older →
Firefly aggregates headlines and links to original sources. All content belongs to its respective publishers.