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 |
|---|---|---|---|---|
hn hn-ai | 3d | Project Copacetic: Directly patch container image vulnerabilities | — | ↗Not briefed |
arxiv arxiv-ai | 3d | Data-Efficient Backdoor Attacks on Code Models via Trigger-Neutralized Counterfactual Scoring | Pre-trained code models are exposed to backdoor poisoning, in which crafted training samples make the model misbehave on a trigger. A… | ↗Not briefed |
arxiv arxiv-ai | 3d | Beyond Final Outcomes: Quantifying Claim-Level Information Dynamics in Multi-Agent LLM Interaction | When LLM-powered agents interact, information is repeatedly shared, expanded, and forgotten, yet multi-agent evaluation typically focu… | ↗Not briefed |
arxiv arxiv-ai | 3d | 4DMemory: Benchmarking Long-Horizon Spatiotemporal Memory with Harness-Augmented Agents | Humans efficiently process multimodal streams spanning hours to years, building persistent, fine-grained spatiotemporal memories that… | ↗Not briefed |
bluesky bluesky-ai | 3d | 😭 | 😭 | ↗Not briefed |
arxiv arxiv-ai | 3d | When to Trust Memory: Retrieval-Guided Probabilistic Spatiotemporal Forecasting under Distribution Shift | Spatiotemporal forecasting plays an important role in real-world applications, where probabilistic approaches provide uncertainty esti… | ↗Not briefed |
arxiv arxiv-ai | 3d | TGR-EvoDAG: A DAG-based Multi-Agent Code Evolution Framework for Generative Recommendation on Tencent | Large Language Model (LLM) agents have enabled automated code evolution for algorithm discovery. However, current recursive self-impro… | ↗Not briefed |
arxiv arxiv-ai | 3d | ChoreoStudio: Planning What, When, and Who in Music-Driven Group Choreography🔧 Modal | Music-driven group dance generation must decide not only how each dancer moves, but also what formation the group should form, when it… | ↗Not briefed |
arxiv arxiv-ai | 3d | From Visual Affinity to Learning Credit in Multimodal Reinforcement Learning | Reinforcement learning with verifiable rewards (RLVR) improves multimodal reasoning, but uniformly assigning a single response-level a… | ↗Not briefed |
arxiv arxiv-ai | 3d | Beyond Gain: Measuring Skill Internalization for Agent Skill Retirement | LLM agents retrieve skills from public hubs that hold up to about 100,000 entries and serve many models. As models improve, a hub must… | ↗Not briefed |
hn hn-ai | 3d | Show HN: Pulse reached 8 contributors und 30 stars | For a while I was dedicated working on SaaS and whatnot. Few months ago I completely took a different turn, and went to the world of F… | ↗Not briefed |
bluesky bluesky-ai | 3d | カバンからキムチのにおい漏れてんなコレ これもLLMのせいですか？ | カバンからキムチのにおい漏れてんなコレ これもLLMのせいですか？ | ↗Not briefed |
arxiv arxiv-ai | 3d | System One models Are Not Better Rerankers | System One models, or decision models, answer a question about an input with a calibrated probability instead of generated text. Withi… | ↗Not briefed |
bluesky bluesky-ai | 3d | Fires are something that can be relatively easily detected from sensor data, so this is one of those times "AI" is likely to actually be us… | Fires are something that can be relatively easily detected from sensor data, so this is one of those times "AI" is likely to actually… | ↗Not briefed |
bluesky bluesky-ai | 3d | "AI" is a really overloaded term. It's possible that some kind of machine learning tech could help predict wildfires, but this might be som… | "AI" is a really overloaded term. It's possible that some kind of machine learning tech could help predict wildfires, but this might b… | ↗Not briefed |
rss bing-news-ai | 3d | Navy memo directs use of ‘super intelligence’ instead of AI | The switch better reflects the Navy’s approach to scaling the technology in its operations at supersonic speed, according to the memo. | ↗Not briefed |
rss bing-news-ai | 3d | Navy memo directs use of ‘super intelligence’ instead of AI | The switch better reflects the Navy’s approach to scaling the technology in its operations at supersonic speed, according to the memo. | ↗Not briefed |
rss bing-news-ai | 3d | Navy memo directs use of ‘super intelligence’ instead of AI | The switch better reflects the Navy’s approach to scaling the technology in its operations at supersonic speed, according to the memo. | ↗Not briefed |
rss bing-news-ai | 3d | Navy memo directs use of ‘super intelligence’ instead of AI | The switch better reflects the Navy’s approach to scaling the technology in its operations at supersonic speed, according to the memo. | ↗Not briefed |
rss bing-news-ai | 3d | Navy memo directs use of ‘super intelligence’ instead of AI | The switch better reflects the Navy’s approach to scaling the technology in its operations at supersonic speed, according to the memo. | ↗Not briefed |
arxiv arxiv-ai | 3d | SKG in Practice: OOD Generalization Cases Across FourDomains | This paper presents the engineering implementation methods of the SKG inference framework and provides complete case studies across fo… | ↗Not briefed |
arxiv arxiv-ai | 3d | GeoLEON-CLIP: Graph-Anchored Multimodal Contrastive Learning for Geographic Region Representation🔧 Modal | Geographic regions are multimodal: satellite imagery captures visual morphology, text conveys human-interpretable semantics, and map s… | ↗Not briefed |
arxiv arxiv-ai | 3d | Supervised Source Separation of Electric Vehicle Interior Noise Using a Hybrid Transformer Architecture | The interior sound of an electric vehicle is a superposition of several distinct noise contributions - wind noise, tire noise, powertr… | ↗Not briefed |
arxiv arxiv-ai | 3d | Many Perturbations, Shared Failure Dynamics: Tracing Direction-Sensitive Perturbation Influence in Language Models | Why can a task-preserving rewrite turn a correct answer into a failure while a larger internal change leaves the answer intact? We dev… | ↗Not briefed |
arxiv arxiv-ai | 3d | Solvent: Long-Horizon Reasoning for Multi-Step Retrosynthesis | Multi-step retrosynthesis requires recursively transforming a target molecule through successive reaction steps into purchasable build… | ↗Not briefed |
hn hn-ai | 3d | Remove pgrust: the entry is cheating | — | ↗Not briefed |
bluesky bluesky-ai | 3d | あと、今まで見過ごされてきた脆弱性が明らかになったということと、杜撰な管理で漏洩したことはまた別のレイヤーの話だと思うので、全てLLMのせいにするのは無理がある気がする。 | あと、今まで見過ごされてきた脆弱性が明らかになったということと、杜撰な管理で漏洩したことはまた別のレイヤーの話だと思うので、全てLLMのせいにするのは無理がある気がする。 | ↗Not briefed |
arxiv arxiv-ai | 3d | DMAIC-Bench: From Self-Reflection to Self-Restraint for Agents in Multi-Stage Analytical Tasks | Multi-stage analytical tasks require LLM agents to use intermediate evidence to decide whether to continue, redirect or stop. Self-ref… | ↗Not briefed |
hn hn-ai | 3d | Toward fast AI interpretability and control | — | ↗Not briefed |
bluesky bluesky-ai | 3d | That's a really broad statement. If a version of that statement is even true, you're saying people need to use neural network or transforme… | That's a really broad statement. If a version of that statement is even true, you're saying people need to use neural network or trans… | ↗Not briefed |
bluesky bluesky-ai | 3d | Yeah I’ll take the other side of this. We are going to have a very bad two years and then the web will be much more stable and secure than … | Yeah I’ll take the other side of this. We are going to have a very bad two years and then the web will be much more stable and secure… | ↗Not briefed |
hn hn-ai | 3d | Nvidia GTC Berlin 2026 | — | ↗Not briefed |
arxiv arxiv-ai | 3d | CoPiT: A conditional Position-induced Transformer for aerodynamic field prediction on unstructured meshes | High-fidelity Reynolds-averaged Navier--Stokes (RANS) simulations remain a standard tool for aerodynamic analysis, yet each solve on a… | ↗Not briefed |
bluesky bluesky-ai | 3d | Switch 2 jailbreak via software-only exploit Nintendo removed SmileBASIC 4 from the eShop after a supposed Switch 2 jailbreak, 'ns2jb', wa… | Switch 2 jailbreak via software-only exploit Nintendo removed SmileBASIC 4 from the eShop after a supposed Switch 2 jailbreak, 'ns2jb… | ↗Not briefed |
arxiv arxiv-ai | 3d | Lifelong Context Adaptation through Bidirectional Memory-Skill Learning | Lifelong LLM agents must learn from an expanding history of interactions while making each decision within a bounded context window. M… | ↗Not briefed |
arxiv arxiv-ai | 3d | Complementing DINO Features with Image Structure for Part Discovery | Self-supervised Vision Transformers (ViTs) like DINO encode rich semantics, enabling training-free segmentation of objects from the ba… | ↗Not briefed |
bluesky bluesky-ai | 3d | URGENT ACTION: An Australian proposed law affects all authors wherever you live. Make your voice heard, repost asap ‘TheWriters Guild’ stat… | URGENT ACTION: An Australian proposed law affects all authors wherever you live. Make your voice heard, repost asap ‘TheWriters Guild’… | ↗Not briefed |
bluesky bluesky-ai | 3d | Faire un POC avec un LLM prend 2 jours. Le mettre en prod sans sueurs froides prend 6 mois 😅 Comment évaluer un modèle imprévisible ? Er… | Faire un POC avec un LLM prend 2 jours. Le mettre en prod sans sueurs froides prend 6 mois 😅 Comment évaluer un modèle imprévisible… | ↗Not briefed |
bluesky bluesky-ai | 3d | "Up to four humans and unlimited agents" So I can just stop at 4 humans and win that money to spend on a non AI project? | "Up to four humans and unlimited agents" So I can just stop at 4 humans and win that money to spend on a non AI project? | ↗Not briefed |
bluesky bluesky-ai | 3d | As a maintainer of a similarly important piece of Internet infrastructure the exploitation line rings true. For a lot of OSS folks we hav… | As a maintainer of a similarly important piece of Internet infrastructure the exploitation line rings true. For a lot of OSS folks w… | ↗Not briefed |
arxiv arxiv-ai | 3d | ReLEAP: Upcycling Vision Datasets through Cross-Task Unification and Semi-Supervised Annotation | Developing robust vision models requires large amounts of annotated data, especially for tasks involving fine-grained class distinctio… | ↗Not briefed |
bluesky bluesky-ai | 3d | My last point is that journalists are repeating OpenAI press releases here that the vulnerability was discovered with AI when in fact if yo… | My last point is that journalists are repeating OpenAI press releases here that the vulnerability was discovered with AI when in fact… | ↗Not briefed |
bluesky bluesky-ai | 3d | I know Bluesky loves their proprietary platforms but like... Kind of a big week OpenResponses, OpenSandbox, AgentSkills, and OpenEnv | I know Bluesky loves their proprietary platforms but like... Kind of a big week OpenResponses, OpenSandbox, AgentSkills, and OpenEnv | ↗Not briefed |
bluesky bluesky-ai | 3d | Es scheitert nach wie vor bereits an der OCR. Und nun kommen noch die bekannten LLM-Probleme hinzu. Meinen damaligen Test des Tools gibt es… | Es scheitert nach wie vor bereits an der OCR. Und nun kommen noch die bekannten LLM-Probleme hinzu. Meinen damaligen Test des Tools gi… | ↗Not briefed |
arxiv arxiv-ai | 3d | Trial by Trace: Benchmarking AI Agent Harnesses for Reliability and Failure Attribution | LLM agents are increasingly used to solve real-world problems, and their effectiveness depends on both the underlying models and the h… | ↗Not briefed |
bluesky bluesky-ai | 3d | All your devices with “always on” front cameras are selling videos of you to Anthropic to teach their humanoids’ facial expressions. Some o… | All your devices with “always on” front cameras are selling videos of you to Anthropic to teach their humanoids’ facial expressions. S… | ↗Not briefed |
hn hn-ai | 3d | Building a cheaper review loop for coding agents using Jev | — | ↗Not briefed |
arxiv arxiv-ai | 3d | "Not" is Not Lost in Translation: A Causal Analysis of Negation Representations in Multilingual LLMs | Negation occurs in every language, yet its surface realization differs drastically. Despite this, multilingual large language models (… | ↗Not briefed |
bluesky bluesky-ai | 3d | I’m trying to make sense of what Ondřej Surý says in this article versus what OpenAI says. Genuinely trying to understand why if they knew … | I’m trying to make sense of what Ondřej Surý says in this article versus what OpenAI says. Genuinely trying to understand why if they… | ↗Not briefed |
arxiv arxiv-ai | 3d | RootDown: Deploying and Benchmarking Models for Beet-Root Detection on Edge Devices | Quality inspection of sugar beets after harvest and storage is critical to preserve sugar content and ensure optimal processing effici… | ↗Not briefed |