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 | BELIEFRAG: Making Adaptive RAG State-Aware under Evolving Evidence | Adaptive RAG uses signals such as confidence, relevance, support, and retrieval quality to decide when to search or correct evidence.… | ↗Not briefed |
arxiv arxiv-ai | 1mo | ID Balancing: Stable Training of Extremely Sparse MoE via PID-Based Load Control | Scaling Large Language Models (LLMs) via Mixture-of-Experts (MoE) enables massive parameter growth with nearly constant per-token comp… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Hank Green suggested creators make AI codes of conduct and I thought that was a nice idea. Sometimes when I end up interacting with an LLM,… | Hank Green suggested creators make AI codes of conduct and I thought that was a nice idea. Sometimes when I end up interacting with an… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Characterizing High Bandwidth Flash for LLM Serving | Large language model (LLM) serving requires substantial memory to store model weights and KV caches. As models grow larger and context… | ↗Not briefed |
bluesky bluesky-ai | 1mo | an artist friend used rovo (atlassian's llm assistant, which she has access to through her job) to write a chrome extension that creates ji… | an artist friend used rovo (atlassian's llm assistant, which she has access to through her job) to write a chrome extension that creat… | ↗Not briefed |
hn hn-ai | 1mo | GLM 5.3 flash on AMD GPUs 670 tok/s | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | CDMD: A Cross-Dataset Mixed-Type Diffusion Model for Tabular Data | Generative models for tabular data are typically trained separately for each dataset, limiting knowledge transfer and requiring the st… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Prequential E-Values for Selected-GP Near-Optimality Certificates | When optimizing an expensive black-box function sequentially, as in hyperparameter optimization, we may want to stop once the best eva… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Diagnosing On-Policy Self-Distillation for Reasoning Language Models | On-policy self-distillation (OPSD) has attracted growing interest as a promising approach to improve the reasoning ability of language… | ↗Not briefed |
arxiv arxiv-ai | 1mo | The Row Normalization Puzzle in Muon | This paper examines how row-wise renormalization affects Muon, focusing on the gap between NorMuon's worst-case guarantees and its pra… | ↗Not briefed |
bluesky bluesky-ai | 1mo | It seems Meta and OpenAI are now in cuteness face off when it comes to AI agents. | It seems Meta and OpenAI are now in cuteness face off when it comes to AI agents. | ↗Not briefed |
hn hn-ai | 1mo | Caisi Cyber Evaluations | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | "There are some parts of life that we cannot and should not automate", says CEO who creates products that automate all parts of life while … | "There are some parts of life that we cannot and should not automate", says CEO who creates products that automate all parts of life w… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Bongard: Training Machine Intuition | Human intelligence relies heavily on learned intuition: recognising patterns and judging situations without explicitly unfolding every… | ↗Not briefed |
arxiv arxiv-ai | 1mo | T-Router: Learning Thalamic Routing for Reasoning with Parameter-Efficient Reinforcement Learning | Parameter-efficient reinforcement learning aims to improve reasoning with a compact trainable interface to a pretrained model. We intr… | ↗Not briefed |
arxiv arxiv-ai | 1mo | MASCRDM: Multi-Agent System for Compliance Risk Detection and Mitigation in Training Process of Large Language Models | Large Language Models (LLMs) have been applied in various fields. However, ensuring compliance and safety of LLMs, such as avoiding di… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "Altman said it was “bad for the world if OpenAI waits too long to go public”"- www.ft.com/content/211d... Beyond parody pomposity. "Ask… | "Altman said it was “bad for the world if OpenAI waits too long to go public”"- www.ft.com/content/211d... Beyond parody pomposity.… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Reserve-Aware Contrast Certificates for Conservative Bandits with Uncertain Baselines | Conservative bandits must improve an incumbent policy without exhausting a prescribed performance budget. When the incumbent is uncert… | ↗Not briefed |
hn hn-ai | 1mo | CAISI's Assessment of Z.ai's GLM-5.3 Cyber Capabilities | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | False Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search Agents | Self-evolving search agents build their own training curricula by jointly optimizing a proposer that generates questions and a solver… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Prediction: Steering VLM Agents with Retrospective World Modeling | Equipping VLM agents with world modeling capabilities has shown strong potential for complex reasoning and long-horizon planning, whil… | ↗Not briefed |
hn hn-ai | 1mo | What do you want from AI? | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Generalisation Signal Need Not Be a Model-Selection Signal | Model selection in computational biology often relies on validation data drawn from the training regime, even when deployment lies out… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Anxiety Checklist – 3 years of itterate with no success yet | I lived with Anxiety Disorders (and Panic Attack, avoidance behavior) from 27 to 37 years old, at first I saw my general practionner w… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Big shades of Microsoft’s infamous Fear, Uncertainty, and Doubt (FUD) campaign against Linux in peak Windows years. Certainly seems like p… | Big shades of Microsoft’s infamous Fear, Uncertainty, and Doubt (FUD) campaign against Linux in peak Windows years. Certainly seems l… | ↗Not briefed |
hn hn-ai | 1mo | Lightweight Resilient Recursive Parsing | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning Infinite-Horizon Average-Reward CMDPs via State Augmentation | We study infinite-horizon average-reward constrained Markov decision processes (CMDPs) under the weakly communicating assumption. Exis… | ↗Not briefed |
rss techmeme | 1mo | Singapore doubles down on chipmaking as the AI boom lifts its economy, but high costs, scarce land and labor, and US trade scrutiny threaten its ambitions (Owen Walker/Financial Times) | Owen Walker / Financial Times: Singapore doubles down on chipmaking as the AI boom lifts its economy, but high costs, scarce land and… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Steepest Guidance: A Practical and Principled Approach to Inference-Time Alignment of Flow and Diffusion-based Models | Inference-time alignment of flow and diffusion-based models is critical for achieving flexible generative modeling. Theoretically, Doo… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A strategic roadmap for an atomistic machine-learning ecosystem | Data-driven machine learning (ML) techniques have become an essential tool in many domains of science. Their application to atomistic… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SCIC: Scope- and Codebook-Aware Instruction Conditioning for Speaker-Adapted Expressive TTS | Long-form live-streaming TTS requires context-dependent prosody and paragraph-level coherence. However, many existing instruction-base… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Trustworthy Runtime Error Healing in Real-World Repositories: A Benchmark and Guardrail | Runtime error healing lets a crashed program continue by generating code that repairs its live runtime state. Recent work shows that L… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Shared Phase and Retention Control for Efficient Adaptive Spectral Recurrence | As new evidence arrives, a sequence model must update what it remembers and how memory influences predictions. While Transformers incu… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The AI industry’s tactic has been “to substitute identifying the problem for actually solving it. But at some point, you cannot blog your w… | The AI industry’s tactic has been “to substitute identifying the problem for actually solving it. But at some point, you cannot blog y… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Coding Agents for Coding Theory | We spent five weeks using an LLM coding agent on open problems in coding theory: finding large sets of four-letter words, such as DNA… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Association profile conditioning in a set-temporal transformer for cross-session intracortical motor decoding | Intracortical motor decoders degrade across sessions because the set of recorded units changes and persisting units can alter how thei… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Parameter symmetries determine representational geometry in overparameterized nonlinear networks | Representations are routinely used across machine learning, psychology, and neuroscience to draw inferences about the computations of… | ↗Not briefed |
rss techmeme | 1mo | Alibaba's ModelScope and OSChina's MoArk are vying to become Chinese alternatives to Hugging Face; ModelScope hosts 170K+ open models, while MoArk hosts ~20K (Viola Zhou/Rest of World)🔧 Hugging Face | Viola Zhou / Rest of World: Alibaba's ModelScope and OSChina's MoArk are vying to become Chinese alternatives to Hugging Face; ModelSc… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Multi-LLM Collaborative Alignment via Stackelberg Games | A pool of language models can collaborate and improve collectively by learning from one another's responses. These interactions depend… | ↗Not briefed |
arxiv arxiv-ai | 1mo | RAGScope: A Leakage-Controlled, Cost-Aware Evidence-Gating Protocol for RAG Hallucination Triage | Retrieval-augmented generation (RAG) systems need inexpensive ways to route generated answers: accept low-risk outputs, review uncerta… | ↗Not briefed |
arxiv arxiv-ai | 1mo | HO-FL: Hybrid-Order Federated Learning for Heterogeneous Edge Devices | Federated learning (FL) on memory-constrained edge devices faces a dilemma: first-order (FO) optimization (i.e., backpropagation) dema… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Text: LLM-Based Dimensional Emotion Evaluation in Multimodal Dialogue | Emotion recognition in conversation has been widely studied, but applying Large Language Models (LLMs) to continuous dimensional emoti… | ↗Not briefed |
arxiv arxiv-ai | 1mo | LexReward: A Taxonomy-Driven Reward Framework for Legal Language Models | Legal language models require reward signals that capture not only answer correctness but also the multidimensional quality of legal r… | ↗Not briefed |
hn hn-ai | 1mo | Repeated VM Escapes by GPT-5.6-Cyber Based Agents🔧 GPT-5 | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | 🚨 Anthropic warned investors in its IPO prospectus that AI could pose “catastrophic or existential risks to humanity” and said its models … | 🚨 Anthropic warned investors in its IPO prospectus that AI could pose “catastrophic or existential risks to humanity” and said its mo… | ↗Not briefed |
hn hn-ai | 1mo | Dirty Optimization Secrets (C for Playdate) | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | CORE: Conflict-Oriented Reasoning Elimination for Verifiable Language-Model Search | Test-time reasoning systems often respond to failure by restarting or revising the latest step, even when an earlier decision caused t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SparseEngine: Sparse-First Inference Engine🔧 vLLM | Long-context LLM agents accumulate interaction histories that strain KV-cache memory and attention computation. Although sparse attent… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Argus: A Real-EKS Study of When Predicting Spot Interruptions Beats Simple Checkpointing | Elastic Compute Cloud (EC2) Spot is 60% to 90% cheaper than On-Demand but can be reclaimed on just a 2-minute notice; for expensive mu… | ↗Not briefed |
rss techmeme | 1mo | Hands-on with Dots, OpenAI's work-focused agentic product: highly capable and intuitive, with natural-feeling conversations represented as a chat inside ChatGPT (Casey Newton/Platformer)🔧 ChatGPT | Casey Newton / Platformer: Hands-on with Dots, OpenAI's work-focused agentic product: highly capable and intuitive, with natural-feeli… | ↗Not briefed |