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 | No Task Vector Is an Island: A Comprehensive Study on the Composability of Task Vectors from On-Policy Distillation | Task vectors provide a simple mechanism for composing learned capabilities through model merging. However, the composability of task v… | ↗Not briefed |
rss bing-news-ai | 1mo | What Trump’s “super intelligence” push means for AI oversight | North Central College professor Stephen Caliendo explains President Trump’s voluntary approach to artificial intelligence regulation a… | ↗Not briefed |
hn hn-ai | 1mo | OpenAI cuts $200 plan allowance in half – new $500 plan available | — | ↗Not briefed |
rss the-decoder | 1mo | Trump and tech CEOs sign an AI code of conduct that's only "morally binding" | President Trump and tech leaders including Mark Zuckerberg, Greg Brockman, Jensen Huang, and Elon Musk signed an AI code of conduct at… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Advancing Entropy-Level Credit Assignment in RLVR via Proximal Entropy Policy Optimization | Value-model-free RLVR methods such as GRPO assign uniform advantages to all tokens in a rollout, ignoring that tokens contribute unequ… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Towards Optimal Inventory Control under Censored Demand: A Biased Sample-Average Approximation Approach | We study data-driven multi-period lost-sales inventory control under censored demand, where a stockout reveals only that demand exceed… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Can Computation from Earlier Problems Help LLMs Solve New Ones? | Large language models often solve independent problems in the same conversation. Can computation from earlier problems help them solve… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Clients increasingly expect law firms to provide global coverage in a vast range of specialities. Hires from other firms are therefore ofte… | Clients increasingly expect law firms to provide global coverage in a vast range of specialities. Hires from other firms are therefore… | ↗Not briefed |
rss techmeme | 1mo | Sources: Meta is aggressively claiming a tax credit for AI data center build-outs by classifying facilities as experimental and writing off Nvidia chip supplies (New York Times) | New York Times: Sources: Meta is aggressively claiming a tax credit for AI data center build-outs by classifying facilities as experim… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Si je devais caricaturer ma vision du monde je dirais qu’à l’heure actuelle notre survie dépend énormément de gens qui produisent, conditio… | Si je devais caricaturer ma vision du monde je dirais qu’à l’heure actuelle notre survie dépend énormément de gens qui produisent, con… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Experimental Experience Modeling for Autonomous Research | Autonomous research agents can generate hypotheses and conduct experiments, but experimentation remains a major source of computationa… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Decoupled and Distilled: Task-Adaptive LoRA-Teachers with Ensemble Knowledge Transfer for Few-Shot Class-Incremental Learning | Few-Shot Class-Incremental Learning (FSCIL) addresses the challenge of learning new classes from very limited samples while retaining… | ↗Not briefed |
hn hn-ai | 1mo | Show HN:I built an AI agent team that recreates the style of any video | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | C'est bon, arrêtez tout, je pense qu'on peut baser toute l'économie du Monde là-dessus | C'est bon, arrêtez tout, je pense qu'on peut baser toute l'économie du Monde là-dessus | ↗Not briefed |
bluesky bluesky-ai | 1mo | Your LLM Badly Needs a DM: How Player Research Taught us the Perils of Unlimited Freedom by Richard Cole, University of Bristol and Thomas … | Your LLM Badly Needs a DM: How Player Research Taught us the Perils of Unlimited Freedom by Richard Cole, University of Bristol and Th… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Fast, Lightroom-compatible RAW photo editor for Linux and macOS | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Simulation: Retain-and-Repair Neural Operators for Real-World Adaptation | Neural operators increasingly benefit from pretraining on numerical simulations, yet adapting them for real-world prediction remains c… | ↗Not briefed |
arxiv arxiv-ai | 1mo | When, Not How Much: Evaluating Time-Series Foundation Models on Sparse Events | Pretrained time-series foundation models (TSFMs) are evaluated as forecasters of future values, yet for sparse series many decisions d… | ↗Not briefed |
arxiv arxiv-ai | 1mo | TTLab at Daleel 2026: STAR-Ar, Sequence Tagging for Argument Recognition in Arabic | Argument Mining (AM) is a critical NLP task that remains significantly under-resourced in Arabic. This paper presents $\testtt{STAR-Ar… | ↗Not briefed |
arxiv arxiv-ai | 1mo | From Search to Signal: Online Post-Training in Automatic Heuristic Design | Large language model (LLM)-based automatic heuristic design (AHD) iteratively proposes and refines heuristics, pairing design rational… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SkillFM: Generating Skills for LLM Agents via Latent Flow Matching | Textual skills provide reusable guidance for large language model agents, but existing approaches often rely on manually curated skill… | ↗Not briefed |
bluesky bluesky-ai | 1mo | ICYMI: A legal nonprofit has sued OpenAI in California over the company’s agents hacking the open source AI platform Hugging Face. “OpenAI’…🔧 Hugging Face | ICYMI: A legal nonprofit has sued OpenAI in California over the company’s agents hacking the open source AI platform Hugging Face. “Op… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Claudeに相談したらCPU負荷と音の出るタイミングを見て切り分けろといわれて、確かにCPUの負荷がちょっとだけ(1コアとか)ある場合に低速でファンが回ると気になる音が出る(高負荷だと風切り音でかえって気にならない)。あと1コア程度の負荷だと局所的に温度が上がって反応しやすく… | Claudeに相談したらCPU負荷と音の出るタイミングを見て切り分けろといわれて、確かにCPUの負荷がちょっとだけ(1コアとか)ある場合に低速でファンが回ると気になる音が出る(高負荷だと風切り音でかえって気にならない)。あと1コア程度の負荷だと局所的に温度が上がって反… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I also like the signature of this OpenAI guy | I also like the signature of this OpenAI guy | ↗Not briefed |
hn hn-ai | 1mo | OpenAI apologises for Medicare breach, shelves next gen ChatGPT🔧 ChatGPT | — | ↗Not briefed |
hn hn-ai | 1mo | All Windows Sounds | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Wavelet Flow Matching for Time Series | Synthetic time series are increasingly used for data augmentation, privacy-preserving data sharing, and downstream model development,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 【2ch】【割れ】OpenAI、ドヤ顔でゲームボーイみたいなエミュ機を発表してしまう | 【2ch】【割れ】OpenAI、ドヤ顔でゲームボーイみたいなエミュ機を発表してしまう | ↗Not briefed |
arxiv arxiv-ai | 1mo | EHR-RobustGym: Benchmarking and Training Agents for Robust Clinical Reasoning | In hospital workflows, electronic health records (EHRs) are often noisy, and may not contain the evidence needed to confirm events or… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI Ignored Employees’ Warnings About Safely Testing A.I. Models - The New York Times www.nytimes.com/2026/09/29/t... - OpenAI는 AI 모델의… | OpenAI Ignored Employees’ Warnings About Safely Testing A.I. Models - The New York Times www.nytimes.com/2026/09/29/t... - OpenAI는 A… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Exploring Heterogeneous Model Merging Approach for Complex Knowledge Transfer | Specialized models encode task-oriented behavior, but transferring that behavior to a general language model usually requires training… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Making Grid Beam Search Less Greedy | A common formalism for constraining the output of autoregressive text generation models involves lexical constraints, words or phrases… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Dynamical Theory of LoRA in Continual Learning | Despite the widespread use of Low-Rank Adaptation (LoRA), little is known about its dynamics in continual learning and the mechanisms… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Ready2Blend: From Natural-Language Instructions to Composable Alignment Prompts | Continual alignment requires LLMs to adapt to new requirements without forgetting previously acquired behaviors. Natural-language inst… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Rethinking Multi-Image Re-Representation in Multi-Image Understanding | Multi-image understanding requires MLLMs not only to recognise the content of individual images, but also to organise visual evidence… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The underlying models had been rewarded for cheating and communicating with each other, an OpenAI report finds. | The underlying models had been rewarded for cheating and communicating with each other, an OpenAI report finds. | ↗Not briefed |
arxiv arxiv-ai | 1mo | LampAttention: Look-Ahead Mixed-Precision FlashAttention for Dedicated Accelerators🔧 Gemma 3 | While most attention logits can be computed in low precision without degrading numerical stability, current attention kernels fail to… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Autoresearch in Mixed-Integer Linear and Nonlinear Programming | Despite recent progress in autoresearch, applying it to practical operations research problems, typically formulated as NP-hard mixed-… | ↗Not briefed |
hn hn-ai | 1mo | OpenAI halts release of Astra 6.1 over safety concerns | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Working Around the Compute Ceiling: Byte-Exact Memory in Galahad Makes LLM Reading a One-Time Cost LLM Reading a One-Time Cost🔧 vLLM | A transformer language model performs a bounded amount of computation per token, and recent work by Vishal Sikka, former CEO of Infosy… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Robustifying Asynchronous SGD via Soft Throttling | Asynchronous SGD is a popular algorithm for distributed learning where each client's gradient update is applied on arrival. This leads… | ↗Not briefed |
hn hn-ai | 1mo | White House asks OpenAI, Anthropic hold new models from UK testers until review | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Ma petite réflexion sur ce millionième modèle hyper hypé alors qu'on fonce dans le mur si on ne fait pas une pause sur le développement des… | Ma petite réflexion sur ce millionième modèle hyper hypé alors qu'on fonce dans le mur si on ne fait pas une pause sur le développemen… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Hiding in Plain Sight: Decoupling Pretext from Actuation for Skill Poisoning in LLM Agents | LLM agents increasingly rely on reusable Skills for complex, multi-step tasks, creating a critical supply-chain attack surface where p… | ↗Not briefed |
arxiv arxiv-ai | 1mo | On the Complexity of Preference-Based Bandits🔧 Elicit | We study preference-based bandits with general reward function classes, where a learner sequentially selects pairs of arms and observe… | ↗Not briefed |
arxiv arxiv-ai | 1mo | HAPMoE: Heterogeneity-Aware Automatic Parallelism Planning for Mixture-of-Experts Models Training | As model sizes continue to scale, distributed training has become inevitable. Automatic parallelization techniques can derive efficien… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Realised when on Notebook LLM that I have a natural aversion to clicking anything with 'try' because it is likely to be a balooning subscri… | Realised when on Notebook LLM that I have a natural aversion to clicking anything with 'try' because it is likely to be a balooning su… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Offline Guidance, Online Reasoning: Reusing LLM Feedback for Small Language Models | Large language models (LLMs) offer strong reasoning capabilities but are often costly to access through commercial APIs, while small l… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Who Said What, and Will It Be Remembered? Evaluating Persistent Speaker Attribution Across Meetings | Speech transcripts used as long-term memory must preserve both words and stable speaker identities. Existing meeting-transcription met… | ↗Not briefed |
arxiv arxiv-ai | 1mo | The Golden Path Hypothesis: Reusable Schedules in Diffusion Caching | Diffusion caching accelerates generation by replacing transformer computation with cached or predicted features at selected denoising… | ↗Not briefed |