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 |
|---|---|---|---|---|
bluesky bluesky-ai | 1mo | That reminds me, I came across a great term this morn on FB where someone in a computer group said ‘Gemini says…<rambling shite>’ and the o… | That reminds me, I came across a great term this morn on FB where someone in a computer group said ‘Gemini says…<rambling shite>’ and… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Joint Precision Neural Networks: Task-Aware Dependency and Predictive Learning | Exploiting meaningful latent structures from data to solve downstream tasks is a fundamental challenge in signal processing and machin… | ↗Not briefed |
bluesky bluesky-ai | 1mo | なんかLLM(AI生成)技術って新技術というか裕福な3代目の過去の親の遺産食い潰しっぽい。食い尽くしたらあと潰れるしかないみたいな。 | なんかLLM(AI生成)技術って新技術というか裕福な3代目の過去の親の遺産食い潰しっぽい。食い尽くしたらあと潰れるしかないみたいな。 | ↗Not briefed |
arxiv arxiv-ai | 1mo | Scalable Minimal-Change Learning for Controllable Image Editing🔧 FLUX | Image editing should change only the attributes specified by an instruction while preserving everything else, yet current methods ofte… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Adaptive Expert Guidance for Efficient On-Policy Reinforcement Learning | With massively parallel simulation, on-policy Reinforcement Learning methods such as PPO have become standard in many domains. However… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Investigating Query-Insensitive Behavior in Spatio-Temporal Video Grounding | Spatio-temporal video grounding (STVG) aims to localize objects or events described by natural language queries in both space and time… | ↗Not briefed |
hn hn-ai | 1mo | İstanbul – Kars Evden Eve Nakliyat | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Parallelism or Concession? Concurrency-Aware Procurement Negotiation for Agentic Commerce | Agentic buyers can cheaply fork a procurement task into many parallel negotiations, but concurrency is not free: every thread consumes… | ↗Not briefed |
bluesky bluesky-ai | 1mo | By build, you mean, waste precious AI credits to have Claude do it for you? | By build, you mean, waste precious AI credits to have Claude do it for you? | ↗Not briefed |
arxiv arxiv-ai | 1mo | Spectral Geometry of Attention: From Information Routing to Uncertainty | In this work, we study transformer attention through the lens of spectral geometry and operator theory. We view each attention head as… | ↗Not briefed |
arxiv arxiv-ai | 1mo | D-Loop: Looped Diffusion Drafting for Speculative Decoding | Block diffusion accelerates speculative decoding by drafting multiple tokens in one forward pass. However, each position predicts a ma… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Ultrasound Operator Guidance Using World Modeling and Retrieval Based Action Planning | Ultrasound is widely used, but acquisition quality is heavily dependent on the operator's knowledge and expertise. With demand for exa… | ↗Not briefed |
hn hn-ai | 1mo | Scalable, Efficient Optical Architecture for Muxed Deepfake Video Detection | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | youtu.be/ibZryrOJdQo?... 対米80兆円投資の大部分は「米国のAI電力」負担に使われているの(・・? | youtu.be/ibZryrOJdQo?... 対米80兆円投資の大部分は「米国のAI電力」負担に使われているの(・・? | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning While Scheduling Jobs under Context-Dependent Service Rates: An Anytime Rate-Optimal Algorithm | We study contextual queueing bandits, where a learner schedules jobs while learning unknown service rates modeled by logistic function… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Elon har skruvat lite på Twitters inbyggda LLM igen. | Elon har skruvat lite på Twitters inbyggda LLM igen. | ↗Not briefed |
bluesky bluesky-ai | 1mo | 「Mythos 級の能力」を持つ Local LLM が出始めたことも、世界中でセキュリティインシデント増加の歯止めがきかなくなってきている原因のひとつであろうと認識しています（いろいろある中の、無視できないひとつ） | 「Mythos 級の能力」を持つ Local LLM が出始めたことも、世界中でセキュリティインシデント増加の歯止めがきかなくなってきている原因のひとつであろうと認識しています（いろいろある中の、無視できないひとつ） | ↗Not briefed |
bluesky bluesky-ai | 1mo | Når en LLM først skulle sende to fagorganiserte til feil adresse, så var det jo for så vidt flaks at de havna hos oss, som kunne gi dem rik… | Når en LLM først skulle sende to fagorganiserte til feil adresse, så var det jo for så vidt flaks at de havna hos oss, som kunne gi de… | ↗Not briefed |
arxiv arxiv-ai | 1mo | P3: Persistent Particle Planning for Constrained Diffusion Control | Diffusion models provide expressive priors over trajectories, but adapting these priors to test-time constraints requires maintaining… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Il faut demander au LLM qui a fait l'affiche (et qui a au moins eu la sincérité de se mettre en premier sur le liste) | Il faut demander au LLM qui a fait l'affiche (et qui a au moins eu la sincérité de se mettre en premier sur le liste) | ↗Not briefed |
arxiv arxiv-ai | 1mo | AgentSpy: Making AI Agent Behavior Observable | AI agents built on large language models (LLMs) run shell commands, read and write files, and reach the network, typically with their… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Langevin Flow Maps: Efficient Molecular Dynamics and Transition Path Sampling | Molecular dynamics simulations proceed by integrating the Langevin equations over many small femtosecond timesteps. This poses a chall… | ↗Not briefed |
hn hn-ai | 1mo | Aegis – Next-Gen Selfhosted Web Application Firewall (WAF) & Edge Reverse Proxy | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Borehole – acquisition tech due diligence from a repo's Git history | I've done technical due diligence on acquisitions for years at Amazon. This is typically a very time consuming process, taking w… | ↗Not briefed |
arxiv arxiv-ai | 1mo | EpicWorldModel: Exploration-driven Planning with Latent World Models | Latent world models based on Joint-Embedding Predictive Architecture (JEPA) are deterministic by design. While successful in fully obs… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: daytab – 1.4M human studies aggregated for interventions and outcomes | this is a project i have worked on for years.<p>i've always found it tedious to do health research for supplements/outcomes… | ↗Not briefed |
arxiv arxiv-ai | 1mo | How (and How Not) to Use Data Augmentation in VLA Post-Training | Vision-language-action (VLA) models currently demonstrate strong performance in a wide range of real-world robotics tasks. However, th… | ↗Not briefed |
rss techmeme | 1mo | An interview with CoreWeave Physical AI SVP Richard Ahlfeld on AI models failing real-world checks, the roles of synthetic data and physical tests, and more (Superintelligence) | Superintelligence: An interview with CoreWeave Physical AI SVP Richard Ahlfeld on AI models failing real-world checks, the roles of sy… | ↗Not briefed |
hn hn-ai | 1mo | Open source didn't get less important because AI can write the code | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Breaking the Tie: A Cluster-Aware Routing Framework for Large Language Models🔧 Consensus | With the rapid development of artificial intelligence, the emergence of various Large Language Models (LLMs) has created a rich model… | ↗Not briefed |
hn hn-ai | 1mo | I don't go to Cornell: Unrelated Asian men targeted in case of mistaken identity | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Massive US AI companies have no viable business model, argues Bruce Schneier. 🤖 The security legend discusses his book Rewiring Democracy… | Massive US AI companies have no viable business model, argues Bruce Schneier. 🤖 The security legend discusses his book Rewiring Demo… | ↗Not briefed |
bluesky bluesky-ai | 1mo | At BookHaven we continue to get requests from companies to buy our secondhand books for LLM training & we continue to refuse those orders. … | At BookHaven we continue to get requests from companies to buy our secondhand books for LLM training & we continue to refuse those ord… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 【漫画家の皆さまへ】 LLM含めて「生成AI完全不使用」の場合は、プロフ欄に「制作過程に生成AI不使用」もしくは「生成AI不使用」と明記するようにお願いします。 海外で「日本の漫画家は生成AI使用する」と伝わっており、信頼に関わる事態となっております。 | 【漫画家の皆さまへ】 LLM含めて「生成AI完全不使用」の場合は、プロフ欄に「制作過程に生成AI不使用」もしくは「生成AI不使用」と明記するようにお願いします。 海外で「日本の漫画家は生成AI使用する」と伝わっており、信頼に関わる事態となっております。 | ↗Not briefed |
arxiv arxiv-ai | 1mo | Byte Language Models: Scaling, Emergent Abstractions, and Information Allocation | Tokenizer-free language models remove the inductive bias of fixed tokenizers by modeling text directly as bytes, but the resulting lon… | ↗Not briefed |
arxiv arxiv-ai | 1mo | StagQ: Constraint-Driven Multi-Precision Weight Quantization for LLMs | Serving a large language model (LLM) across a fleet of deployments requires several weight-precision operating points. Multi-precision… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Can Language Models Learn to Reject Their Own Bad Reasoning Steps? | Verifier-guided decoding can prevent harmful reasoning steps from contaminating subsequent generation, but typically relies on an exte… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Transfer-Stratified On-Policy Distillation for RL-Improved Reasoning Teachers | Reinforcement learning can substantially improve a reasoning teacher, but it is unclear which of those improvements survive when the t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Reachability-Aware Diffusion Policy Optimization | Diffusion policies provide expressive action distributions for continuous-control reinforcement learning. However, safety-aware online… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Fast Last-Iterate Convergence in Zero-Sum Markov Games with Bandit Feedback | We study last-iterate convergence in unknown two-player zero-sum discounted Markov games with bandit feedback. The players learn indep… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: The Velocity Trap, the Phoenix Project for the AI Age | I'm taking a bit of a risk with this one, this is a novella I authored with the help of AI. If that offends you, please move on!<… | ↗Not briefed |
bluesky bluesky-ai | 1mo | #andrewcampbell #britishartist #art #subvert #meta #agitate #algorithms #provoke #X #attack #openai #resist #ai #disturb #instagram #artfor… | #andrewcampbell #britishartist #art #subvert #meta #agitate #algorithms #provoke #X #attack #openai #resist #ai #disturb #instagram #a… | ↗Not briefed |
arxiv arxiv-ai | 1mo | JLD: Perceptual Distance Through A Jacobian Lens | Image compression, restoration, and generation all require a way to measure how different two images look to a person. Pixel error ign… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Mais @maliciarogue.bsky.social a fait remarquer que 17 était souvent proposé spontanément par les LLM | Mais @maliciarogue.bsky.social a fait remarquer que 17 était souvent proposé spontanément par les LLM | ↗Not briefed |
bluesky bluesky-ai | 1mo | Level up your game AI! 🚀 Explore Laya, the open-source decision model that runs LOCALLY on players' PCs. Fast, free, and perfect for offli… | Level up your game AI! 🚀 Explore Laya, the open-source decision model that runs LOCALLY on players' PCs. Fast, free, and perfect for… | ↗Not briefed |
arxiv arxiv-ai | 1mo | HuatuoGPT-3: RL-Only Domain Adaptation from Base Models | Domain adaptation aims to turn a general-purpose large language model (LLM) into an expert for a target domain. While the dominant SFT… | ↗Not briefed |
bluesky bluesky-ai | 1mo | In addition to the fact of every corpus of every LLM containing the whole works of Graeme Bird, there's also the known fact know that Graem… | In addition to the fact of every corpus of every LLM containing the whole works of Graeme Bird, there's also the known fact know that… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Strategic Multi-Agent Learning for Interpretable Action Valuation of All Players in Football | Valuing player actions in football requires accounting for strategic interactions among 22 players, including off-ball movements and d… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Physics-Informed but Not Physics-Consistent: Error Geometry and Subspace Projection for Neural AC Power Flow | Recent neural power-flow solvers, including emerging foundation models, achieve accurate voltage predictions, yet such accuracy does n… | ↗Not briefed |
rss marktechpost | 1mo | Yandex Introduces Sona: A Single Generative Recommender That Replaces Entire Recommendation Cascade | One transformer ran candidate generation and ranking in Yandex Music's A/B test without hand-engineered features, lifting likes 11.42%… | ↗Not briefed |