News
The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.
| Source | Age | Title | Summary | Status |
|---|---|---|---|---|
arxiv arxiv-ai | 1mo | eval-unlearn: Benchmarking unlearning in Text-to-Image Diffusion Models | The rising number of concept unlearning techniques for text-to-image (T2I) diffusion models has produced a fragmented evaluation lands… | ↗Not briefed |
bluesky bluesky-ai | 1mo | It is a survey paper, not a study with original experiments. It organizes existing research on multi-agent systems involving multiple coope… | It is a survey paper, not a study with original experiments. It organizes existing research on multi-agent systems involving multiple… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SpikeCredit: Temporal Credit Carrier for Reinforcement Learning with Sparse Rewards | Reinforcement learning (RL) with sparse rewards is challenging because delayed outcomes provide little guidance about which intermedia… | ↗Not briefed |
rss techmeme | 1mo | An OpenAI agent security executive discusses the Hugging Face incident, OpenAI's response, sandboxing improvements, alignment, "reasonable paranoia", and more (Joe/@joedaroo)🔧 Hugging Face | Joe / @joedaroo: An OpenAI agent security executive discusses the Hugging Face incident, OpenAI's response, sandboxing improvements, a… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Decide – Jev decisions in the shell, scripts, and agent skills🔧 OpenRouter | All software makes decisions. Code handles the deterministic ones. Decision models like Jev handle the judgement calls.<p>A decision m… | ↗Not briefed |
bluesky bluesky-ai | 1mo | NYC Council wants a kill switch + third-party validation on every AI system sold in the city, $25K/violation, plus a whistleblower bounty. … | NYC Council wants a kill switch + third-party validation on every AI system sold in the city, $25K/violation, plus a whistleblower bou… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Florida asks court to bar OpenAI from developing new models without oversight as part of child harm lawsuit - court filing | Florida asks court to bar OpenAI from developing new models without oversight as part of child harm lawsuit - court filing | ↗Not briefed |
bluesky bluesky-ai | 1mo | FLORIDA ASKS COURT TO BAR OPENAI FROM DEVELOPING NEW MODELS WITHOUT OVERSIGHT AS PART OF CHILD HARM LAWSUIT -- COURT FILING | FLORIDA ASKS COURT TO BAR OPENAI FROM DEVELOPING NEW MODELS WITHOUT OVERSIGHT AS PART OF CHILD HARM LAWSUIT -- COURT FILING | ↗Not briefed |
bluesky bluesky-ai | 1mo | ChatGPT Grok Claude geminiに課金しちゃってるんだけど、誰落としていいと思う？ https://dtsoku.com/chatgpt-grok-claude-gemini%e3%81%ab%e8%aa%b2%e9%87%91%e3%81%97%e3%8…🔧 ChatGPT | ChatGPT Grok Claude geminiに課金しちゃってるんだけど、誰落としていいと思う？ https://dtsoku.com/chatgpt-grok-claude-gemini%e3%81%ab%e8%aa%b2%e9%87%91%e3%81%97… | ↗Not briefed |
arxiv arxiv-ai | 1mo | WavePP: High-Throughput Pipeline Parallel LLM Prefill under Prefix Reuse🔧 vLLM | Pipeline parallelism can improve prefill throughput by processing multiple request chunks concurrently across different stages of the… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Rubric-Aware On-Policy Self-Distillation for LLM Personalization | LLM personalization aims to generate responses aligned with individual users' preferences and needs. User-specific rubrics make these… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Imprint Reader: From Weight-Update Readout to Behavioral Intervention🔧 Elicit | As language models take a growing role in AI development, a natural aspiration is for them to reflect on their own learning process, a… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Latency and accuracy tradeoffs in Spiking Neural Networks | Spiking neural networks are attractive for low-power speech command recognition, yet their latency has received far less attention tha… | ↗Not briefed |
arxiv arxiv-ai | 1mo | On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics | On-policy learning has been argued to reduce catastrophic forgetting, produce sparser parameter updates, and improve generalisation. H… | ↗Not briefed |
arxiv arxiv-ai | 1mo | AIM-ZO: Activation-Informed Subspace Maintenance for Zeroth-Order LLM Fine-Tuning | Zeroth-order (ZO) optimization offers a memory-efficient alternative for LLM fine-tuning by estimating updates only from forward evalu… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Towards Reliable AI Data Scientists: Data Agents with Workflow Harnesses | Large language model agents are increasingly deployed for data-intensive work, yet reliable data analysis requires more than general-p… | ↗Not briefed |
hn hn-ai | 1mo | Senzii – open-source staff scheduling with a native MCP interface | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | SCBO: Semantically Coherent Batching and Ordering for LLM-Based Social Surveys | Large Language Models (LLMs) offer a scalable way to simulate survey respondents using demographic profiles and observed reference res… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Spatial Grafting: Grounding 3D Features for Flow-Matching Robot Policies | Pretrained robot manipulation policies such as vision-language-action models (VLAs) or world-action models (WAMs) leave interaction-re… | ↗Not briefed |
arxiv arxiv-ai | 1mo | AnswerMap: Faithful Spatial Interpretability of VLMs from Answer Posteriors | When a VLM answers a visual query, current interpretability tools rely on text rationales, which use a mismatched modality, or on inte… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: SQL Based Memory for LLMs | — | ↗Not briefed |
rss techmeme | 1mo | AI research leaders at OpenAI, Anthropic, Microsoft, and Meta warn of an impending "intelligence explosion" and call for oversight into automated AI research (Maxwell Zeff/Wall Street Journal) | Maxwell Zeff / Wall Street Journal: AI research leaders at OpenAI, Anthropic, Microsoft, and Meta warn of an impending “intelligence e… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Disentangling Lung-Cancer CT/LDCT AI: A Systematic Evidence Map of Clinical Tasks, Evidence Chains, and Translational Gaps | Artificial-intelligence studies using computed tomography (CT) for lung cancer are often broadly labelled "prediction" despite address… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 1. Anthropic hired Accenture as an AI watchdog. Critics call it a conflict of interest. 2. An AI bot made a false nuclear-weapons report t… | 1. Anthropic hired Accenture as an AI watchdog. Critics call it a conflict of interest. 2. An AI bot made a false nuclear-weapons rep… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Long-Horizon Scaling: How Model Capabilities Shape the Returns to Computation | Long-horizon agents improve solutions through sustained interaction, execution, and task feedback. Scaling studies relate performance… | ↗Not briefed |
bluesky bluesky-ai | 1mo | O primeiro-ministro Anthony Albanese afirmou que as investigações continuam e que a Austrália manifestou diretamente ao CEO da OpenAI, Sam … | O primeiro-ministro Anthony Albanese afirmou que as investigações continuam e que a Austrália manifestou diretamente ao CEO da OpenAI,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 📢 Agente de IA da OpenAI invadiu portal governamental de dados de saúde da Austrália A Austrália revelou que um agente de IA desenvolvido… | 📢 Agente de IA da OpenAI invadiu portal governamental de dados de saúde da Austrália A Austrália revelou que um agente de IA desenvo… | ↗Not briefed |
arxiv arxiv-ai | 1mo | EP-Mem: Elastic Privacy Memory for Social Relationship-Aware LLM Agents | Large language model (LLM) agents face critical privacy risks when acting as delegates in human-agent-human communication. To prevent… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Selection: Token Parameterization for Extreme Visual Token Compression🔧 Modal | Visual-token compression is effective for improving the efficiency of vision-language models, but under extreme compression budgets, t… | ↗Not briefed |
hn hn-ai | 1mo | GLM 5.3 signed my commit as Claude Fable 5. The harness is in the weights | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Token-Disentangled Latent Test-Time Scaling for Vision-Language Reasoning | Latent test-time scaling improves reasoning by refining hidden states during inference, but existing methods typically apply a single… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: A NaN-immune explicit ODE solver that survives where RK4 explodes | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Generative AI-Based Data Augmentation for Oral Lesion Classification: The PhotoMOCI Dataset and Benchmark | Early detection of oral cancer via photographic imaging presents a promising avenue for large-scale oral cavity screening. However, th… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SignFLIP: A Unified Model for Sign Language Translation and Generation via Stage-wise Alignment at Scale | Sign language translation and generation share the goal of bidirectional alignment between text and sign representations. However, exi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | TANGO: Watermarking Masked Diffusion Language Models in Token Pairs | Masked-diffusion language models fill in masked positions in parallel and in no fixed order. Most practical text watermarks assume lef… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This absolutely cannot be good. I don’t think that the demented, greedy, stupid, corrupt fool understands or cares about the destructive po… | This absolutely cannot be good. I don’t think that the demented, greedy, stupid, corrupt fool understands or cares about the destructi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | No matter what hypothetical good use cases there are for LLM text, the sloppers will always outnumber the thoughtful users. We are already … | No matter what hypothetical good use cases there are for LLM text, the sloppers will always outnumber the thoughtful users. We are alr… | ↗Not briefed |
bluesky bluesky-ai | 1mo | i know this feeling! it's funny talking to an llm about my plans and it is like "18 months and 5 people, sir" and then i'm like "it's just … | i know this feeling! it's funny talking to an llm about my plans and it is like "18 months and 5 people, sir" and then i'm like "it's… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Temporal Heterogeneous Graph Pretraining for Relational Deep Learning | Relational deep learning models database rows and foreign-key links as a heterogeneous graph for prediction from record attributes and… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Top AI researchers at OpenAI, Anthropic, Microsoft and Meta are warning that AI could soon begin improving itself faster than humans can un… | Top AI researchers at OpenAI, Anthropic, Microsoft and Meta are warning that AI could soon begin improving itself faster than humans c… | ↗Not briefed |
rss techmeme | 1mo | SpaceX launches its enormous Starship rocket into orbit for the first time, deploying 26 of the most advanced Starlink satellites to join the 11,000 in service (Marcia Dunn/Associated Press) | Marcia Dunn / Associated Press: SpaceX launches its enormous Starship rocket into orbit for the first time, deploying 26 of the most a… | ↗Not briefed |
bluesky bluesky-ai | 1mo | God I hate that LLM way of “talking”, it’s like seeing a dog crawl on its belly on the floor bsky.app/profile/wolv... | God I hate that LLM way of “talking”, it’s like seeing a dog crawl on its belly on the floor bsky.app/profile/wolv... | ↗Not briefed |
rss techcrunch-ai | 1mo | Modulate raises $25M for its voice models and analysis suite | Modulate deploys its models to detect deepfake, fraud and scam | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Hierarchy of Entropy-Shapley Games for Multivariate Predictive Uncertainty | Modern probabilistic machine learning models increasingly produce multivariate outputs with complex dependence structure, from multi-s… | ↗Not briefed |
arxiv arxiv-ai | 1mo | ASCT: Attentive Search over Counterfactual Trees for Credit Assignment in Agentic Reinforcement Learning | Terminal utility evaluates a complete agentic workflow, but learning requires credit for the decisions within it. We introduce Attenti… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Jensen Huang says AI distillation is 'competition.' Scott Bessent has called it 'theft' | Jensen Huang says AI distillation is 'competition.' Scott Bessent has called it 'theft' | ↗Not briefed |
arxiv arxiv-ai | 1mo | Uncertainty Quantification in Cardiac Model Personalisation from Ultrafast Ultrasound | Cardiac model personalisation requires inferring mechanical parameters that are not directly measurable in vivo. Ultrafast ultrasound… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Adversarial Consistency-Guided Representation Learning for Multi-view Clustering | Multi-view clustering aims to capture cross-view consistency while exploiting view-specific information. However, shared representatio… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Understanding On-Policy Distillation: A Mechanistic Interpretability Perspective via Sparse Crosscoders | On-policy distillation (OPD) is a widely adopted post-training technique for LLM reasoning. It is commonly believed to transfer knowle… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "There are plenty of reasons to believe that the ‘hyperscaler’ model of AI can never be profitable." I largely agree with that. www.thene… | "There are plenty of reasons to believe that the ‘hyperscaler’ model of AI can never be profitable." I largely agree with that. www.… | ↗Not briefed |