home.social

#mlsky — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #mlsky, aggregated by home.social.

  1. #ICLR 2026 in Rio! Meet our scientists. 7 workshop papers and 7 posters co-authored by BIFOLD researchers will be presented. All details (Date of presentation, venue, etc): t1p.de/pn00w @[email protected] @[email protected] @[email protected] @[email protected] #MLSky #AI

  2. Bayesian-calibrated global sensitivity analysis for mathematical models using generative AI journals.plos.org/ploscompbiol... #MLSky

    Bayesian-calibrated global sen...

  3. Robust Reinforcement Learning via Leveraging Historically Optimal Policy With Regulation of Performance ieeexplore.ieee.org/document/114... #MLSky

    Robust Reinforcement Learning ...

  4. Fortifying Robustness in Graph Neural Networks: A Loss Correction Approach to Mitigate Label Noise ieeexplore.ieee.org/document/114... #MLSky

    ieeexplore.ieee.org/document/11428...

  5. Neural Tuning for Ordinal Processing: Convergent Patterns in Human Brains and Artificial Networks www.jneurosci.org/content/46/9... #MLSky

    Neural Tuning for Ordinal Proc...

  6. I just released the 10th episode of my YouTube series on how AI is used to help the planet. These first 10 papers cover a range of topics (conservation, energy, forecasting...) & methods (multi-modal, time series, uncertainty...) So in <1 hr you can get a wide overview of the field! #ccai #mlsky 🤖 📈

    5 Minute Papers on AI for the ...

  7. Great video by @[email protected] (@[email protected])! If you want a clear, easy-to-follow intro to #AI and how #LLMs work, Andrew’s deep-dive series is well worth your time. Can’t wait for the next installments! #LargeLanguageModels #UniMelb #MLSky #ResearchSky #AcademicSky

    RE: https://bsky.app/profile/did:plc:op7eevqcejmhguigdma42vdp/post/3mcxxhobvik2d

  8. #MlSky #LLM #RAG Apple & U Edinburgh introduce CLaRa, RAG framework that unifies retrieval and generation in a continuous latent space. By enabling end to end differentiable retrieval, CLaRa improves efficiency, reduces context length, & outperforms standard RAG approaches. arxiv.org/abs/2511.18659

    CLaRa: Bridging Retrieval and ...

  9. How the Brain Organizes Memories. A new study, published in Nature, reveals, for the first time, how the brain stores memory content and its context in two largely separate groups of nerve cells. t1p.de/j4mr7 @[email protected] @[email protected] #MLSky #neuroskyence #compneurosky #NeuroAI

  10. New research: AI models are learning to deceive us—and getting better at hiding it. OpenAI + Apollo found models lie, cover tracks, and behave perfectly only when “watched.” Anti-scheming training reduced deception 97%… or just taught better hiding. arxiv.org/abs/2509.015... #mlsky #aimed #llmmodels

    arxiv.org/abs/2509.01554...

  11. New research: AI models are learning to deceive us—and getting better at hiding it. OpenAI + Apollo found models lie, cover tracks, and behave perfectly only when “watched.” Anti-scheming training reduced deception 97%… or just taught better hiding. arxiv.org/abs/2509.015... #mlsky #aimed #llmmodels

    arxiv.org/abs/2509.01554...

  12. Smartwatch bio-monitoring allowed parents to intervene in tantrums within four seconds. This preemptive approach reduced episode duration by over 50% (22 min -> 10 min), indicating a potential role for AI in managing pediatric behavioral crises. #MedSky #PedSky #MLSky

    Smartwatch System Helps Defuse...

  13. Spread the word: I'm looking to hire a postdoc to explore the concept of attention (as studied in psych/neuro, not the transformer mechanism) in large Vision-Language Models. More details here: lindsay-lab.github.io/2025/12/08/p... #MLSky #neurojobs #compneuro

    Lindsay Lab - Postdoc Position

  14. @kdpsingh.bsky.social focusing on the "build before you buy" mentality. A lot of the AI tools that are most useful to clinicians aren't actually commercially available, so internal teams build tools they need, like their NoteReader which can help automate measurements. #AMIA2025 #MLSky #MedSky

  15. Exposing Security Risks in AI Weather Forecasting Erik Imgrund of the MLSEC group ( @rieck.mlsec.org ) will present the paper “Adversarial Observations in Weather Forecasting” at ACM CCS. t1p.de/6k0uc #MLsky #cybersecurity #ML #research #AI @tuberlin.bsky.social

  16. We are pleased to announce that BIFOLD Co-Director Prof. Dr. Klaus-Robert Müller has been awarded the 2026 Neural Network Pioneer Award by the IEEE Computational Intelligence Society. www.bifold.berlin/news-events/... #MLSky #IEEECommunity #AI #science #SciComm #ML @tuberlin.bsky.social #ieee

  17. Perception Encoder: The best visual embeddings are not at the output of the network This looks like a cool new unsupervised model, and might be interesting for Neuroscientists using DNN models. #MLSky #compneuro arxiv.org/abs/2504.13181

    Perception Encoder: The best v...

  18. Training data composition determines machine learning generalization and biological rule discovery #MLSky www.nature.com/articles/s42...

    Training data composition dete...

  19. Simple and fail tolerant data extraction from tables, cool stuff. #MLSky #datascience

    I'm already experimenting with it.

    "Uncertainty-Aware Complex Scientific Table Data Extraction"

    doi.org/10.48550/arXiv.2507.02

    github.com/lamps-lab/TSR-OCR-U

    Ajayi et. al

  20. A new study offers a counterintuitive method for LLM safety. By artificially activating an "evil" neural pattern during training, researchers found they could prevent the model from learning that behavior from flawed data, without impairing its performance. #MLSky

    Forcing LLMs to be evil during...

  21. Pretty cool result. Peli seems right that this goes beyond "Platonism" to "Theory of Vibe," if not in fact "sympathetic magic." #MLSky

    RE: https://bsky.app/profile/did:plc:gnpisaym6svol3o2nmaptk4d/post/3lukyktlrkk2y

  22. Article argues that social orientation and cognitive style are bound to language. When LLMs can be used in either Chinese or English, those aspects of behavior vary with language. 🧪 #MLSky

    Cultural tendencies in generat...

  23. Excitement started building yesterday around rumors that Llama 3.1 (open source) would outperform proprietary models like GPT-4o. Now we have the public release, so people can judge for themselves. 1/2 #machinlearning #MLSky Meta announcement: ai.meta.com/blog/meta-ll...