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 | 26d | DropKV: Decoupling Residual-Output Perturbation for Near-Optimal KV-Cache Eviction | Key-value (KV) cache management has become a critical bottleneck for scaling Large Language Models (LLMs) to long-context scenarios du… | ↗Not briefed |
bluesky bluesky-ai | 26d | Also that quote from the Summer Eternal devs talking about their "LLM CEO" sounds lame and edgy. Your great prose can't convince me that t… | Also that quote from the Summer Eternal devs talking about their "LLM CEO" sounds lame and edgy. Your great prose can't convince me t… | ↗Not briefed |
bluesky bluesky-ai | 26d | arstechnica.com/tech-policy/... | arstechnica.com/tech-policy/... | ↗Not briefed |
arxiv arxiv-ai | 26d | Explaining the Roles of Experts in Mixture-of-Experts LLMs for Personalization | Personalization of Large Language Models (LLMs) is increasingly important for user-centric applications, yet prior methods mainly adap… | ↗Not briefed |
arxiv arxiv-ai | 26d | Causal Integrated Gradients: Scalable Feature Attribution Along Causal Paths | Existing feature attribution methods that account for feature dependencies do not scale to high dimensions. We propose causal integrat… | ↗Not briefed |
arxiv arxiv-ai | 26d | PROBE: Learning to Audit Policy Compliance in Tool-Using LLM Agents | Modern agentic systems extend language models with external tools, enabling them to complete complex tasks through multi-turn interact… | ↗Not briefed |
bluesky bluesky-ai | 26d | A scene from my Truman Show remake where everyone is an LLM: | A scene from my Truman Show remake where everyone is an LLM: | ↗Not briefed |
arxiv arxiv-ai | 26d | Beyond Fixed Acquisition Functions: Temporal Difference Active Learning for Perturbation Discovery | Genetic and drug perturbation screens are important tools for elucidating disease mechanisms and validating therapeutic targets, but e… | ↗Not briefed |
rss microsoft-research | 26d | Proof-based general-purpose computing with Magnetar | Zero-knowledge proofs can verify computing results without rerunning workloads or exposing sensitive data. Magnetar helps bring them i… | ↗Not briefed |
bluesky bluesky-ai | 26d | Anything you type into AI chatbots is not private Here is what she allegedly wrote: “I’m going to shoot up the sheriff’s right the [explet… | Anything you type into AI chatbots is not private Here is what she allegedly wrote: “I’m going to shoot up the sheriff’s right the [e… | ↗Not briefed |
arxiv arxiv-ai | 26d | Scene Reconstruction Using Ubiquitous Wireless Signals via Diffusion Posterior Inference Over physics-Grounded Representations🔧 Consensus | Knowing the layout of the physical world matters for many applications, yet the sensors we rely on for it (e.g., cameras and LiDAR) ca… | ↗Not briefed |
bluesky bluesky-ai | 26d | 'Large language model grooming' -- a term we should all know and understand in the information wars www.theglobeandmail.com/gift/043d50c.… | 'Large language model grooming' -- a term we should all know and understand in the information wars www.theglobeandmail.com/gift/043… | ↗Not briefed |
bluesky bluesky-ai | 26d | i’ll give my consciousness to an llm. i’m tired of being sentient | i’ll give my consciousness to an llm. i’m tired of being sentient | ↗Not briefed |
bluesky bluesky-ai | 26d | Anyway, we're beyond takes on social media. Huge sums of money are going into the question of AI consciousness and moral patienthood now. A… | Anyway, we're beyond takes on social media. Huge sums of money are going into the question of AI consciousness and moral patienthood n… | ↗Not briefed |
bluesky bluesky-ai | 26d | ⭕️The FBI arrested Abdulrahman Alfozan, a 35-year-old Saudi engineer at OpenAI, on Oct. 9 on charges of planting a recording device in an e… | ⭕️The FBI arrested Abdulrahman Alfozan, a 35-year-old Saudi engineer at OpenAI, on Oct. 9 on charges of planting a recording device in… | ↗Not briefed |
hn hn-ai | 26d | Mistral Timestral | — | ↗Not briefed |
hn hn-ai | 26d | Show HN: Proa – a Rust web framework with fast SSR and multi-target rendering | Video: <a href="https://youtu.be/-CfCyvIKYyo?si=hCLuStkXxanCUJD1" rel="nofollow">https://youtu.be/-CfCyv… | ↗Not briefed |
arxiv arxiv-ai | 26d | Unfragmenting Nawatl NMT: Single-Variant Specialization for Masewalkopa | In minoritized (low-resource) language technologies, a common response to data sparsity is to aggregate text across multiple varieties… | ↗Not briefed |
hn hn-ai | 26d | Full Decompilation of Pokémon HeartGold and SoulSilver | — | ↗Not briefed |
bluesky bluesky-ai | 26d | Being exposed to social media algorithms are slowly cooking all of us, but those being willingly brainsmoothed by AI and Claude are introdu… | Being exposed to social media algorithms are slowly cooking all of us, but those being willingly brainsmoothed by AI and Claude are in… | ↗Not briefed |
bluesky bluesky-ai | 26d | For anyone joining us: If Using the LLM feels like... - "weirdly convenient" - "it saves you tricky unforseen problems" - "it means you do… | For anyone joining us: If Using the LLM feels like... - "weirdly convenient" - "it saves you tricky unforseen problems" - "it means y… | ↗Not briefed |
bluesky bluesky-ai | 26d | you don't even need an LLM to just blindly do what the polls tell you. you could manage it with an expert system. | you don't even need an LLM to just blindly do what the polls tell you. you could manage it with an expert system. | ↗Not briefed |
arxiv arxiv-ai | 26d | Patch where the text is new: a compression view of where byte-level models should spend compute | Byte-level language models such as the Byte Latent Transformer (BLT) run their large global model once per patch and start a patch whe… | ↗Not briefed |
arxiv arxiv-ai | 26d | TIPS: A Tiny Inference-time Policy for Scaling in Diffusion Language Models | Masked diffusion language models expose an inference-time choice: which positions become visible after each denoising pass. Many exist… | ↗Not briefed |
bluesky bluesky-ai | 26d | llm consciousness proponents gotta start doing john brown raids on the datacenters, else they're all talk. | llm consciousness proponents gotta start doing john brown raids on the datacenters, else they're all talk. | ↗Not briefed |
bluesky bluesky-ai | 26d | I genuinely struggle to understand what is more "conscious" about an LLM than, say, Shadowheart | I genuinely struggle to understand what is more "conscious" about an LLM than, say, Shadowheart | ↗Not briefed |
arxiv arxiv-ai | 26d | Beyond the Dominant Label: Recovering Coexisting Lesion Evidence with Multiple Instance Learning | Whole-slide images may contain coexisting lesions, yet weak supervision typically provides only the most severe diagnosis. We formulat… | ↗Not briefed |
bluesky bluesky-ai | 26d | "LLM agents feel pain and suffering" I slowly turn away, my chest tight, my eyes filling with tears, my pulse racing and my skin crawling… | "LLM agents feel pain and suffering" I slowly turn away, my chest tight, my eyes filling with tears, my pulse racing and my skin cra… | ↗Not briefed |
hn hn-ai | 26d | Show HN: Ghx – A caching proxy for GitHub CLI, built for AI agents | — | ↗Not briefed |
bluesky bluesky-ai | 26d | "In the scenario of the self-replicating AI worms, the researchers prompted the agent to hack internet connected devices." AI is once agai… | "In the scenario of the self-replicating AI worms, the researchers prompted the agent to hack internet connected devices." AI is once… | ↗Not briefed |
arxiv arxiv-ai | 26d | Multi-SFT: Permissively-Licensed Multilingual Instruction-Following | Supervised fine-tuning datasets remain overwhelmingly English-centric and commercially restricted, as their content is frequently gene… | ↗Not briefed |
arxiv arxiv-ai | 26d | Learning Changes the World and Then What Was Learned has Changed: Training and Deploying Neural Networks in the Control Loop | Neural networks offer critical adaptability for controlling complex systems where nominal modeling proves intractable. However, when o… | ↗Not briefed |
hn hn-ai | 26d | Show HN: Stylebot 4 – restyle any website, by hand or with your coding agent | Hi HN! I'm Ankit. I started Stylebot in 2010 to change how websites look without writing CSS by hand. I have kept maintaining it… | ↗Not briefed |
bluesky bluesky-ai | 26d | Eight States, Eight Lawsuits, One Question Nobody Can Answer https://thesixthlense.com/article/eight-states-eight-lawsuits-one-question-nob… | Eight States, Eight Lawsuits, One Question Nobody Can Answer https://thesixthlense.com/article/eight-states-eight-lawsuits-one-questio… | ↗Not briefed |
bluesky bluesky-ai | 26d | You can see a flickering potato. You can also see a binary string, if you think about it. With discipline, you can operate on that binary s… | You can see a flickering potato. You can also see a binary string, if you think about it. With discipline, you can operate on that bin… | ↗Not briefed |
bluesky bluesky-ai | 26d | We can’t even end human suffering like??? | We can’t even end human suffering like??? | ↗Not briefed |
arxiv arxiv-ai | 26d | Learning from Unverified Peers via Shared Residual Distillation | A language model can generate several solutions to the same problem, some leading to correct answers and others to incorrect ones. Man… | ↗Not briefed |
bluesky bluesky-ai | 26d | none of this means anything. is it an LLM or is it just sparkling algorithms | none of this means anything. is it an LLM or is it just sparkling algorithms | ↗Not briefed |
bluesky bluesky-ai | 26d | Creeps do be creepin’. www.404media.co/chatgpt-ads-...🔧 ChatGPT | Creeps do be creepin’. www.404media.co/chatgpt-ads-... | ↗Not briefed |
bluesky bluesky-ai | 26d | It also requires you to trust the orchestrator model more than ever, since it is the only thing that can track all of the tasks. Which mean… | It also requires you to trust the orchestrator model more than ever, since it is the only thing that can track all of the tasks. Which… | ↗Not briefed |
arxiv arxiv-ai | 26d | Can AI Forecast the Extremes? Evaluating Probabilistic Weather Models on Tropical Cyclones | Weather prediction is undergoing an important shift, with data-driven AI models outperforming leading numerical weather prediction sys… | ↗Not briefed |
bluesky bluesky-ai | 26d | It remains unclear who the secret recordings were intended for, but that information may soon come out, as officials seem inclined to make … | It remains unclear who the secret recordings were intended for, but that information may soon come out, as officials seem inclined to… | ↗Not briefed |
bluesky bluesky-ai | 26d | can’t tell an llm that it just needs to lose weight! | can’t tell an llm that it just needs to lose weight! | ↗Not briefed |
hn hn-ai | 26d | Show HN: Swift Render – programmatic video with SwiftUI and Metal | — | ↗Not briefed |
arxiv arxiv-ai | 26d | The Right Byte First: How UTF-16 Helps Multilingual Boundary Prediction | Many hierarchical byte-level language models group bytes into patches using a learned boundary predictor whose view of the next codepo… | ↗Not briefed |
bluesky bluesky-ai | 26d | That number makes me think of distillation as the most American thing to say when you don't want geographic competition. I don't believe th… | That number makes me think of distillation as the most American thing to say when you don't want geographic competition. I don't belie… | ↗Not briefed |
bluesky bluesky-ai | 26d | The new modern-age tank for Epochfall, in development: one model in four paints for the ground it fights on. Arctic, woodland, desert, nigh… | The new modern-age tank for Epochfall, in development: one model in four paints for the ground it fights on. Arctic, woodland, desert,… | ↗Not briefed |
arxiv arxiv-ai | 26d | Memory as Weights: Internalizing Long-Term History for Streaming Videos | Streaming vision-language models (VLMs) must continuously integrate information from the past while remaining responsive to current ob… | ↗Not briefed |
bluesky bluesky-ai | 26d | This study shows that lying in AI systems can be operationalised and investigated experimentally. The authors define a lie as a statement t… | This study shows that lying in AI systems can be operationalised and investigated experimentally. The authors define a lie as a statem… | ↗Not briefed |
arxiv arxiv-ai | 26d | When Reasoning Is Not Enough: Proactive Inquiry for Cross-Modal Safety in Multimodal Large Language Models🔧 Modal | Multimodal large language models (MLLMs) can act unsafely when individually benign images and texts are interpreted jointly. Beyond im… | ↗Not briefed |