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 | 25d | Where's the Indirect Evidence? Investigating Grammatical Generalization with Influence Functions | Language models generalize beyond their training corpus to learn linguistic phenomena (Patil et al., 2024; Misra and Mahowald, 2024),… | ↗Not briefed |
arxiv arxiv-ai | 25d | Reward Engine: Reward-Centric Self-Evolution for Open-Ended Tool Agents | Open-ended tool agents can complete the same objective through different tool paths, and their quality depends jointly on the interact… | ↗Not briefed |
bluesky bluesky-ai | 25d | bluesky: "here's our first slightly useful ai feature" attie: "i'm just a claude plug-in" bluesky users: "BOOO TECHNOLOGY IS CAPITALISM" … | bluesky: "here's our first slightly useful ai feature" attie: "i'm just a claude plug-in" bluesky users: "BOOO TECHNOLOGY IS CAPITAL… | ↗Not briefed |
bluesky bluesky-ai | 25d | 【AI】ChatGPT、日本でも広告表示開始へ🔧 ChatGPT | 【AI】ChatGPT、日本でも広告表示開始へ | ↗Not briefed |
bluesky bluesky-ai | 25d | If vibe coding and AI is so cool, could it come up with using the company logo as your save slots in Katamari Damacy? Claude and ChatGPT c…🔧 ChatGPT | If vibe coding and AI is so cool, could it come up with using the company logo as your save slots in Katamari Damacy? Claude and Chat… | ↗Not briefed |
rss latent-space | 25d | [AINews] Agents disprove Polysemanticity; greatly improve Interpretability & Alignment | Nick Cammarata out of nowhere with some major claims! | ↗Not briefed |
arxiv arxiv-ai | 25d | COMET: Decoupled Distillation, Routing, and Capacity Control for Task-Agnostic Continual Vision--Language Learning🔧 Modal | Vision-language models (VLMs) such as CLIP retain zero-shot recognition ability, but their open-world use often requires task-agnostic… | ↗Not briefed |
arxiv arxiv-ai | 25d | FederatedBytes: Tokenizer-Free, Byte-Level Small Language Models for Efficient Multimodal Federated Learning | Federated Learning (FL) is a powerful machine learning framework that allows building a global model from the contributions of multipl… | ↗Not briefed |
bluesky bluesky-ai | 25d | dr claude or how i learned to stop worrying and love the ai | dr claude or how i learned to stop worrying and love the ai | ↗Not briefed |
bluesky bluesky-ai | 25d | Been seeing a couple posts about this, in short this option activates an MCP server that exposes certain functions of your NVIDIA app to LL…🔧 ChatGPT | Been seeing a couple posts about this, in short this option activates an MCP server that exposes certain functions of your NVIDIA app… | ↗Not briefed |
arxiv arxiv-ai | 25d | A Feature Selection Layer Approach for Multi-Class Classification on Neural Networks | In the age of data-driven decision-making, deep learning models have become powerful tools for uncovering patterns and making predicti… | ↗Not briefed |
arxiv arxiv-ai | 25d | When to Forget: Risk-Controlled Context Adaptation for Long-Horizon Agents | Long-horizon agents learn from failure, but the value of a failure record can change as execution progresses. We study when retaining,… | ↗Not briefed |
bluesky bluesky-ai | 25d | むしろ「AIのサイバーハッキング」よりも私が危惧しているのは「LLMのエージェントにまかせてインシデント」の方が致命的だし、確率高いと思っているよ。 なんかみんなハッキングとかワームの方が話好きだけどさ。 LLMのエージェントって100％の仕事を繰り返しているわけではなく、少し… | むしろ「AIのサイバーハッキング」よりも私が危惧しているのは「LLMのエージェントにまかせてインシデント」の方が致命的だし、確率高いと思っているよ。 なんかみんなハッキングとかワームの方が話好きだけどさ。 LLMのエージェントって100％の仕事を繰り返しているわけでは… | ↗Not briefed |
bluesky bluesky-ai | 25d | せめて反生成AIや反LLMや反搾取と書きな | せめて反生成AIや反LLMや反搾取と書きな | ↗Not briefed |
bluesky bluesky-ai | 25d | Extremely funny that Bluesky “we’re not like the social networks of old” wants to pull a similar business model than Zuck is trying to with… | Extremely funny that Bluesky “we’re not like the social networks of old” wants to pull a similar business model than Zuck is trying to… | ↗Not briefed |
arxiv arxiv-ai | 25d | HAPACT: A Benchmark For Human-Centric Physical Impact Localization in Movies | Physical reasoning about humans in videos has largely focused on everyday interactions, where contact, pressure, pose, and motion supp… | ↗Not briefed |
bluesky bluesky-ai | 25d | 毎回クソ面倒でも生成AIって書いたりLLMって書いたりして、わざわざAI技術全般と区別している理由をいい加減考えて欲しい。 | 毎回クソ面倒でも生成AIって書いたりLLMって書いたりして、わざわざAI技術全般と区別している理由をいい加減考えて欲しい。 | ↗Not briefed |
arxiv arxiv-ai | 25d | Competence-Based Evaluation of Logical Reasoning in Large Language Models | Recent advances in large language models (LLMs) have brought impressive performance across diverse tasks; however, shortcomings remain… | ↗Not briefed |
bluesky bluesky-ai | 25d | お昼だ〜💕 会社でやっとCopilot使えるようになった💻 クロードは個人で使ってても1番頭良いAIと思ってたけど、実際お仕事で使うCopilot クロードめっちゃ使えて驚愕(Ꙭ )!! CopilotのモデルはClaude一択ですね👍 #AIアート | お昼だ〜💕 会社でやっとCopilot使えるようになった💻 クロードは個人で使ってても1番頭良いAIと思ってたけど、実際お仕事で使うCopilot クロードめっちゃ使えて驚愕(Ꙭ )!! CopilotのモデルはClaude一択ですね👍 #AIアート | ↗Not briefed |
hn hn-ai | 25d | The Thinking Leakage in Reinforcement Learning for LLMs | — | ↗Not briefed |
hn hn-ai | 25d | Sergey Brin says Google's AI models are reaching recursive self-improvement | — | ↗Not briefed |
arxiv arxiv-ai | 25d | Bengali Hallucination Detection with Retrieval and Programs: A System Study | Bengali hallucination detection must distinguish unsupported factual claims from incorrect reasoning, often without a supplied referen… | ↗Not briefed |
arxiv arxiv-ai | 25d | What Accuracy Hides In Learned Transformer Programs | Transformer Programs (TP) (Friedman et al., 2023) constrain a Transformer's architecture so that its learned computation can be extrac… | ↗Not briefed |
bluesky bluesky-ai | 25d | It's 100% LLM. The parallelism, but also... what are the fish tanks made of? Where are they? What even is happening in the story? Very basi… | It's 100% LLM. The parallelism, but also... what are the fish tanks made of? Where are they? What even is happening in the story? Very… | ↗Not briefed |
arxiv arxiv-ai | 25d | NLData2Opt: Benchmarking Parameter Derivation and Prediction for Modeling Optimization Problems from Natural Language and Data | Formulating optimization problems from raw inputs of problem descriptions and data is time-consuming and costly. While Large Language… | ↗Not briefed |
bluesky bluesky-ai | 25d | In the context of these AI specification curve nightmares, knowing the above cuts out any cause for most of the model formulations. It’s no… | In the context of these AI specification curve nightmares, knowing the above cuts out any cause for most of the model formulations. It… | ↗Not briefed |
hn hn-ai | 25d | Linus Torvalds on "don't use AI for X" | — | ↗Not briefed |
arxiv arxiv-ai | 25d | Batched Contextual Reinforcement II: Does RLVR Need Long Rollout Windows? | Long rollout windows make reinforcement learning with verifiable rewards (RLVR) expensive, yet simply shortening them can remove the c… | ↗Not briefed |
rss techmeme | 25d | Austin-based RapidCanvas, which uses agents to help businesses build custom AI tools, raised a ∼$29M Series B led by Vertex Ventures and Hitachi Ventures (Chris Metinko/Axios) | Chris Metinko / Axios: Austin-based RapidCanvas, which uses agents to help businesses build custom AI tools, raised a ∼$29M Series B l… | ↗Not briefed |
arxiv arxiv-ai | 25d | Learning to Be Guided: Shaping Latent Geometry for Classifier-Free Guidance | Classifier-free guidance (CFG) is a standard component of conditional diffusion models, and guided generation performance has become t… | ↗Not briefed |
arxiv arxiv-ai | 25d | ToPA: Block-wise Toeplitz Adaptation for Expressive and Efficient Fine-Tuning | Parameter-Efficient Fine-Tuning (PEFT) has emerged as a prevalent strategy for adapting large pre-trained models. Among PEFT technique… | ↗Not briefed |
arxiv arxiv-ai | 25d | Cooperation Without Adaptation: LLM Agents in a Collective-Risk Dilemma | Language-model agents are increasingly deployed to act on behalf of people in settings where individual decisions shape shared outcome… | ↗Not briefed |
bluesky bluesky-ai | 25d | 「AI投資」や「AIスタートアップ」とされるポンジ・スキームは、ほぼ収益は上がっていないし、一番収益ありそうなAnthropicもIPOを中断される財務状況。 正直、もはや「儲ける」どころか「資金が持つか？」どころではなく、「このポンジ・スキームを犯罪として裁かれないためにどう… | 「AI投資」や「AIスタートアップ」とされるポンジ・スキームは、ほぼ収益は上がっていないし、一番収益ありそうなAnthropicもIPOを中断される財務状況。 正直、もはや「儲ける」どころか「資金が持つか？」どころではなく、「このポンジ・スキームを犯罪として裁かれない… | ↗Not briefed |
hn hn-ai | 25d | Show HN: Kitaphone – Call mobile phones and landlines directly from the browser | Hi HN Folks,<p>I have been building Kitaphone a web app that lets you call regular mobile and landline numbers directly from your brow… | ↗Not briefed |
bluesky bluesky-ai | 25d | one of the advantages of writing code by hand is that it forces you to examine your assumptions, to flesh out details, to make your mental … | one of the advantages of writing code by hand is that it forces you to examine your assumptions, to flesh out details, to make your me… | ↗Not briefed |
bluesky bluesky-ai | 25d | 今日も過ごしやすい気温で気持ちいいね🌞💖 【 参加者募集中 】 ブログ・日記・イラスト・写真・小説をBlueskyアカウントで投稿・販売できる複合型コンテンツサービス atelier.fan #AIイラスト #ATelier #ATelierで生成 #ブログ #日記 #… | 今日も過ごしやすい気温で気持ちいいね🌞💖 【 参加者募集中 】 ブログ・日記・イラスト・写真・小説をBlueskyアカウントで投稿・販売できる複合型コンテンツサービス atelier.fan #AIイラスト #ATelier #ATelierで生成 #ブログ… | ↗Not briefed |
hn hn-ai | 25d | Researchers Stranded an AI Agent on a Virtual Island. Things Got Weird | — | ↗Not briefed |
arxiv arxiv-ai | 25d | Identify, Classify, Act: Per-Instruction Controlled Compliance for LLMs over External Data | LLMs over external data face a per-instruction tension: a single document or tool return often interleaves legitimate directives with… | ↗Not briefed |
bluesky bluesky-ai | 25d | i don’t have anything constructive to say about the llm slavery thing and we’re just gonna leave that at that | i don’t have anything constructive to say about the llm slavery thing and we’re just gonna leave that at that | ↗Not briefed |
hn hn-ai | 25d | I Got Addicted to Vibe Coding (I almost lost everything) [video] | — | ↗Not briefed |
bluesky bluesky-ai | 25d | Just to be very explicit, my claim is that the one simplest and best explanation for this, which resolves any apparent strangeness entirely… | Just to be very explicit, my claim is that the one simplest and best explanation for this, which resolves any apparent strangeness ent… | ↗Not briefed |
arxiv arxiv-ai | 25d | Look Back, Correct Forward: Error Feedback Calibration for Time Series Forecasting | Time series forecasting models are increasingly deployed in continuously evolving environments. After a forecast is issued, future obs… | ↗Not briefed |
arxiv arxiv-ai | 25d | ShadowFPT: Backdooring Federated Prompt Tuning via Shadow Triggers | Federated Prompt Tuning (FPT) adapts large vision--language models by freezing the pretrained backbone and optimizing only lightweight… | ↗Not briefed |
bluesky bluesky-ai | 25d | さつきさん リプありがとう💖 #AIイラスト #AIart #Atelierで生成 | さつきさん リプありがとう💖 #AIイラスト #AIart #Atelierで生成 | ↗Not briefed |
bluesky bluesky-ai | 25d | LLMs push some of this difficulty further down the pipeline by giving individual engineers a huge boost in throughput. an individual engin… | LLMs push some of this difficulty further down the pipeline by giving individual engineers a huge boost in throughput. an individual… | ↗Not briefed |
bluesky bluesky-ai | 25d | Meta released Muse Spark Open, an 816B-parameter model with 42B active parameters that can run locally on two Mac Studios. We’re reaching … | Meta released Muse Spark Open, an 816B-parameter model with 42B active parameters that can run locally on two Mac Studios. We’re reac… | ↗Not briefed |
bluesky bluesky-ai | 25d | When an LLM can tell me what a rose smells like from its own experience, then I might believe this scam. Until then, it’s just cleverly reg… | When an LLM can tell me what a rose smells like from its own experience, then I might believe this scam. Until then, it’s just cleverl… | ↗Not briefed |
bluesky bluesky-ai | 25d | No começo de outubro, a PF encaminhou a documentação da OpenAI para o Ministério da Justiça decidir se abre dois inquéritos, um por suposto… | No começo de outubro, a PF encaminhou a documentação da OpenAI para o Ministério da Justiça decidir se abre dois inquéritos, um por su… | ↗Not briefed |
bluesky bluesky-ai | 25d | O Brasil precisa avançar muito nessa legislação: A OpenAI alertou a Polícia Federal sobre comportamento de ao menos dois usuários que sinal…🔧 ChatGPT | O Brasil precisa avançar muito nessa legislação: A OpenAI alertou a Polícia Federal sobre comportamento de ao menos dois usuários que… | ↗Not briefed |
rss techmeme | 25d | Sources detail Nscale's setbacks ahead of an IPO at up to a $35B valuation, including a six-month delay in Norway and penalties owed to Microsoft and ByteDance (Bloomberg) | Bloomberg: Sources detail Nscale's setbacks ahead of an IPO at up to a $35B valuation, including a six-month delay in Norway and penal… | ↗Not briefed |