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 | A Broader Look at Model Merging: Rethinking Implicit Regularization Induced by Task Arithmetic | Model merging aims to build a multi-task model cheaply by combining the weights of individual task-specific models. To perform well ac… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I'd recommend an LLM workflow for prototyping, planning, static code analysis and 'boring' automation work on the fuzzy edge of a project (… | I'd recommend an LLM workflow for prototyping, planning, static code analysis and 'boring' automation work on the fuzzy edge of a proj… | ↗Not briefed |
arxiv arxiv-ai | 1mo | VisionWeave: Weaving Elastic Visual Representations as a Native Capability of MLLMs | Multimodal large language models have become the dominant paradigm for visual understanding, but incur substantial costs by encoding i… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Decide Before You Look: Learning Which Retrieved Memories Deserve Pixels | Multimodal assistants answer questions from long-term memories that contain images. After retrieval, each retrieved image reaches the… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Do Higher-Order Models Win for Higher-Order Reasons? Rethinking Performance Gains in Hypergraph Learning | Higher-order models (e.g., hypergraph neural networks) often outperform lower-order baselines on hypergraph learning benchmarks, and t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | An LLM will never know that you thanked it. It will never know that you swore at it. It will never know that it made a mistake. It will nev… | An LLM will never know that you thanked it. It will never know that you swore at it. It will never know that it made a mistake. It wil… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning from Revision Consequences: Hindsight Meta-Experience Distillation for Self-Improving Agents | As agents continuously improve by generating and revising Skills, the process that discovers and refines those Skills becomes a learna… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning a Ranking from Human Feedback in Log-Concave Random Utility Models | We study the problem of recovering the ranking of a fixed set of items according to their unknown numerical utilities. At each interac… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Mistral CEO says new AI model beats Chinese ones in some areas reut.rs/4zh8BT8 | Mistral CEO says new AI model beats Chinese ones in some areas reut.rs/4zh8BT8 | ↗Not briefed |
arxiv arxiv-ai | 1mo | Can Agents Work for Everyone? Cross-User Reliability for Mobile GUI Agents in Personalized User Interfaces | Mobile GUI agents increasingly operate on interfaces influenced by users' histories and preferences, but their reliability across diff… | ↗Not briefed |
bluesky bluesky-ai | 1mo | It's very *weird* looking at ai coding session as you switch model mid work and it just assumes all that happen prior is its own thoughts … | It's very *weird* looking at ai coding session as you switch model mid work and it just assumes all that happen prior is its own thou… | ↗Not briefed |
hn hn-ai | 1mo | Buscalid – Paste your site, get a real company that fits as a customer | — | ↗Not briefed |
hn hn-ai | 1mo | Rill: A browser for using Claude Code and Codex from any page | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | 「VLMは勝手に進化しない」 コレが常識みたいになってる。 何故か？ 「ある時点の常識」がVLMだから。 これの一つの解がGoogle方式のVLM自体を極小化してデータベースを付けちゃう形にして、外部的な要因でVLMをバージョンアップする。 コレをLLMに干渉させてリアルタイ… | 「VLMは勝手に進化しない」 コレが常識みたいになってる。 何故か？ 「ある時点の常識」がVLMだから。 これの一つの解がGoogle方式のVLM自体を極小化してデータベースを付けちゃう形にして、外部的な要因でVLMをバージョンアップする。 コレをLLMに干渉させて… | ↗Not briefed |
hn hn-ai | 1mo | Doom, simulated and rendered inside the Firebird SQL database | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | DecepEval: A Benchmark for Evaluating Deception in LLM Agents | As large language model (LLM) agents become increasingly autonomous, they may pursue task performance through deception, raising conce… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic is making a dangerous mistake by training Claude to simulate a conscious entity. It's a machine. The AI mustn't be trained to "un… | Anthropic is making a dangerous mistake by training Claude to simulate a conscious entity. It's a machine. The AI mustn't be trained t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 例えばバイブコーディングに知性・知能の有無は関係ないと思うし。心療内科や小児科（こちらは救急外来）のサポートや臨床予測アルゴリズムツール、あるいは観察研究のサポートに知性、知能は必要とされていないし。人手不足などからCLSに求められるかも、しれないけど。極めて専門性が高くなるか… | 例えばバイブコーディングに知性・知能の有無は関係ないと思うし。心療内科や小児科（こちらは救急外来）のサポートや臨床予測アルゴリズムツール、あるいは観察研究のサポートに知性、知能は必要とされていないし。人手不足などからCLSに求められるかも、しれないけど。極めて専門性が… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Feature Encoding in VAE-based Audio Decoders: Effects of Input, Depth and Distribution | Neural audio synthesis models like the Realtime Audio Variational autoEncoder (RAVE) achieve impressive genera tion quality, yet how t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | ReGraph: A Computational Account of Emergent Generalization in the "what" and "where" Dual Visual Streams | Where generalization capacity--the ability to extract context-invariant relational structures--first emerges remains a central questio… | ↗Not briefed |
hn hn-ai | 1mo | We built a job queue on an LSM tree. This is what we learnt | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | the 'uwu smol bean rogue agents' schtick is wearing thin, i see www.ft.com/content/a5ca... | the 'uwu smol bean rogue agents' schtick is wearing thin, i see www.ft.com/content/a5ca... | ↗Not briefed |
bluesky bluesky-ai | 1mo | LLMの性能に触れるにしても、LLMの知性・知能の有無はあなたたちエンドユーザーが気にしても仕方ないし、LLMの評価に繋がらないものだよって失礼だけど思う | LLMの性能に触れるにしても、LLMの知性・知能の有無はあなたたちエンドユーザーが気にしても仕方ないし、LLMの評価に繋がらないものだよって失礼だけど思う | ↗Not briefed |
arxiv arxiv-ai | 1mo | Confidence Reasoning Graphs: Structured Confidence Estimation for LLM Agents | When using an LLM agent in a consequential domain, making an informed decision about whether to trust its output or intervene requires… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Adapting Vision-Language-Action Models to Unknown Visual Disruptions During Execution | Visual disruptions can arise while a robot is executing a task, leaving a vision-language-action (VLA) policy to respond without knowi… | ↗Not briefed |
rss techmeme | 1mo | South Korean officials are probing whether AI agents were used in recent bank hacks, after President Lee Jae Myung said "signs" suggest AI models were involved (New York Times) | New York Times: South Korean officials are probing whether AI agents were used in recent bank hacks, after President Lee Jae Myung sai… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Hybrid Latent Attention for Looped Language Models | Looped language models apply the same stack of layers T times to each token, which deepens the model without adding parameters but mul… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Been thinking about this a lot. An LLM is in many ways a compendium of solved human problems mediated by a probabilistic lookup function, b… | Been thinking about this a lot. An LLM is in many ways a compendium of solved human problems mediated by a probabilistic lookup functi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Generalized Matheron Variational Implicit Processes | Implicit-process priors specify distributions over functions through sample-forward mechanisms such as Bayesian neural networks and st… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Leveraging a four-quadrant approach for evaluating Redpine Science | Redpine Science gives models and agents a single access point to a wide range of peer-reviewed literature, queried directly through th… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Pseudowords as probes: Large Language Models show little of the sublexical sensitivity that governs human pseudoword processing | Systematicity, the probabilistic mapping of form to meaning, permeates language at all levels, and sublexical cues have been shown to… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SIGMA: Self-Improving Alignment Generalization from a Model Spec | LLM agents are increasingly capable of executing complex tasks and of recursively improving themselves on easy-to-verify objectives su… | ↗Not briefed |
bluesky bluesky-ai | 1mo | SemiAnalysis found Anthropic gives subscribers ~5x more AI usage than OpenAI for the same price. Anthropic is subsidizing subscriptions to… | SemiAnalysis found Anthropic gives subscribers ~5x more AI usage than OpenAI for the same price. Anthropic is subsidizing subscriptio… | ↗Not briefed |
bluesky bluesky-ai | 1mo | あなたは非主流派だからと言われたら、それまでなんだけど。それでも人は頭のなかでずっと反復学習可能で、その精度さえあげていけるのに対して、「学習するものが尽きた」として別の情報を永続的に与え続けなければ精度の向上が期待できないLLMは、膨大な情報と処理能力をもつ極めて精緻なアルゴ… | あなたは非主流派だからと言われたら、それまでなんだけど。それでも人は頭のなかでずっと反復学習可能で、その精度さえあげていけるのに対して、「学習するものが尽きた」として別の情報を永続的に与え続けなければ精度の向上が期待できないLLMは、膨大な情報と処理能力をもつ極めて精… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 「詐欺師どもの看板を磨くな」 今の日本のテックジャーナリストの大部分に言えることは、これに尽きる。 少しコンピューターサイエンスとLLMの仕組みを調べれば「これは人工知能ではないし、実はテック大手も知能だと思っていないし、AGI(汎用人工知能)が発生せずに投資が燃え尽きようとし… | 「詐欺師どもの看板を磨くな」 今の日本のテックジャーナリストの大部分に言えることは、これに尽きる。 少しコンピューターサイエンスとLLMの仕組みを調べれば「これは人工知能ではないし、実はテック大手も知能だと思っていないし、AGI(汎用人工知能)が発生せずに投資が燃え尽… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Dynamic Alignment and Calibration for Multimodal Learning🔧 Modal | Dynamic multimodal learning aims to learn robust representations by adaptively modeling information discrepancies across modalities. H… | ↗Not briefed |
rss techmeme | 1mo | ChatGPT is not only using New Yorker cartoonists' style but also adding their signatures to the fake cartoons it creates, which have gone viral on social media (Andrew Deck/Nieman Lab)🔧 ChatGPT | Andrew Deck / Nieman Lab: ChatGPT is not only using New Yorker cartoonists' style but also adding their signatures to the fake cartoon… | ↗Not briefed |
rss bing-news-ai | 1mo | ARUP Artificial Intelligence Project Selected for Redtail AI Factory Challenge Project | ARUP Laboratories has been selected for the second cohort of challenge projects in Utah's Redtail AI Factory with its project, "Scalin… | ↗Not briefed |
rss bing-news-ai | 1mo | ARUP Artificial Intelligence Project Selected for Redtail AI Factory Challenge Project | ARUP Laboratories has been selected for the second cohort of challenge projects in Utah's Redtail AI Factory with its project, "Scalin… | ↗Not briefed |
rss bing-news-ai | 1mo | ARUP Artificial Intelligence Project Selected for Redtail AI Factory Challenge Project | ARUP Laboratories has been selected for the second cohort of challenge projects in Utah's Redtail AI Factory with its project, "Scalin… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I know using vscode and seeing LLM stuff is a "garbage found in trash can" moment but has no one who supervises this stuff ever seen the co… | I know using vscode and seeing LLM stuff is a "garbage found in trash can" moment but has no one who supervises this stuff ever seen t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 「AI」という誤用はジャーナリストから捨てるべきだ｜地雷魚 note.com/jiraygyo/n/n... 大手の新聞社と通信社が表記を変えれば、地方紙もネットメディアも追随し、やがて官公庁の文書も広告代理店の宣伝文句も変わらざるを得なくなる。 それをしないまま「わかり… | 「AI」という誤用はジャーナリストから捨てるべきだ｜地雷魚 note.com/jiraygyo/n/n... 大手の新聞社と通信社が表記を変えれば、地方紙もネットメディアも追随し、やがて官公庁の文書も広告代理店の宣伝文句も変わらざるを得なくなる。 それをしないま… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Good! Get on with it! www.ft.com/content/a5ca... | Good! Get on with it! www.ft.com/content/a5ca... | ↗Not briefed |
hn hn-ai | 1mo | Faster Software Linking with Mold | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Unfortunately, this is only going to pick up pace from here onwards... I plan at some point on picking apart one of these Claude or whateve… | Unfortunately, this is only going to pick up pace from here onwards... I plan at some point on picking apart one of these Claude or wh… | ↗Not briefed |
bluesky bluesky-ai | 1mo | A New Type of LLM on the Block: Decision-Making Models | A New Type of LLM on the Block: Decision-Making Models | ↗Not briefed |
bluesky bluesky-ai | 1mo | LLMの意思の有無に関しては言語野の回路みたいな情報スキームがつくられてるなーって観測できてるけど、それが意思ないし知性、知能の存在証明には繋がらないし。そもそも入力→出力の一方通行かつレイヤー構造だしLLMくん。人は双方向のループ回路で絶え間なく予測と修正のフィードバックを続… | LLMの意思の有無に関しては言語野の回路みたいな情報スキームがつくられてるなーって観測できてるけど、それが意思ないし知性、知能の存在証明には繋がらないし。そもそも入力→出力の一方通行かつレイヤー構造だしLLMくん。人は双方向のループ回路で絶え間なく予測と修正のフィード… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 現行のLLMだので『人間なら休まないといけない行為を休まずに電気がある限り続けられるぜ！』みたいな技術によるサイバー攻撃を振り回されて企業が被害を受けたりしとる中で、「個人で個人情報管理に気をつけるのも大事！」とか政府に言われてんのホント、不利益被ってる主権者を守るのが政府の仕… | 現行のLLMだので『人間なら休まないといけない行為を休まずに電気がある限り続けられるぜ！』みたいな技術によるサイバー攻撃を振り回されて企業が被害を受けたりしとる中で、「個人で個人情報管理に気をつけるのも大事！」とか政府に言われてんのホント、不利益被ってる主権者を守るの… | ↗Not briefed |
bluesky bluesky-ai | 1mo | using vscode and tab completion appears to do LLM stuff where it finished a line and added a comment "from me" reading 'is this OK? I think… | using vscode and tab completion appears to do LLM stuff where it finished a line and added a comment "from me" reading 'is this OK? I… | ↗Not briefed |
bluesky bluesky-ai | 1mo | LLM以前にAIは内部動作がブラックボックス故に出力結果を説明出来ないし機械学習済みモデルは人が読めない欠点があるのに普及を目指している時点で責任を取る気無いでしょ | LLM以前にAIは内部動作がブラックボックス故に出力結果を説明出来ないし機械学習済みモデルは人が読めない欠点があるのに普及を目指している時点で責任を取る気無いでしょ | ↗Not briefed |