Italiano

Post con #NeSy

#NeSy

Post pubblici su X che usano #NeSy.

Hashtag correlati

  1. antonio vergari ⚔️@tetraduzione
    in #NeSy #AI we know #unargmaxable outputs that cannot be predicted because we force them to have 0 probability as they violate constraints. In this #AAAI24 paper we generalize this: outputs with non-zero probability can still never be predicted 👀
    Andreas Grivas @andreasgrv
    However, when the number of features is less than the number of outputs, you are guaranteed to have many label assignments that can never be predicted irrespective of the input. We call such outputs *unargmaxable*. (4/7)
    27244.2K
  2. Andreas Grivas@andreasgrv
    @andreasgrv
    However, when the number of features is less than the number of outputs, you are guaranteed to have many label assignments that can never be predicted irrespective of the input. We call such outputs *unargmaxable*. (4/7)
    Scarica
    1074.9K
  3. Repubblica✓@repubblica
    La strada è ancora lunghissima, e Nessy Guerra è ancora bloccata in Egitto, impossibilitata non solo a lasciare il Paese, ma pure a uscire di casa. Però l’arresto del suo ex, il cittadino italo-egiziano Tamer Haumada, per lei è uno spiraglio, un raggio di luce dopo anni di buio. Tanto che su Instagram la giovane sanremese condannata dalle autorità egiziane in primo e secondo grado per adulterio, nel dare la notizia la definisce «un enorme passo avanti». L'articolo completo di Marco Lignana su Repubblica Genova #rep
    Scarica
    6186.2K
  4. antonio vergari ⚔️@tetraduzione
    we had today @looselycorrect at @InfAtEd giving us a great talk on "The blessing and curse of Concepts in #XAI and #NeSy integration" spanning: - symbols - explanations - interactions - reasoning - shortcuts - alignment - causality - representation - learning !!!
    Scarica
    26403.8K
  5. Sara Magliacane hiring PhDs in Saarland@saramagliacane
    #XAI, #neurosymbolic methods #nesy and #causal #representation #learning #CRL care about learning #interpretable #concepts, but in different ways. We are organizing this @iclr_conf workshop to bring these communities together & learn from each other 🦾🔥💥 Deadline: 30 Jan 2026
    UCRL Workshop @ ICLR 2026 @UcrlW_iclr2026
    📢 Announcing the Workshop on **Unifying Concept Representation Learning** at ICLR’26 ( @iclr_conf ). When? 26 or 27 April 2026 Where? Rio de Janeiro, Brazil Call for papers, schedule, invited speakers & more: Looking forward to your submissions!
    05161.9K
  6. UCRL Workshop @ ICLR 2026@UcrlW_iclr2026
    📢 Announcing the Workshop on **Unifying Concept Representation Learning** at ICLR’26 ( @iclr_conf ). When? 26 or 27 April 2026 Where? Rio de Janeiro, Brazil Call for papers, schedule, invited speakers & more: Looking forward to your submissions!
    182713.2K
  7. Baskin Engineering at UCSC@BaskinEng
    💡We hosted the 2025 conference on #Neurosymbolic Learning and Reasoning (#NeSy), the longest-standing gathering to discuss research in neurosymbolic #AI! “It's the first time the conference was held in the U.S. as a full conference rather than a workshop." -Prof Leilani Gilpin.
    Scarica
    003104
  9. Samuele Bortolotti@samubortolotti
    @samubortolotti
    🌐 rsbench allows you to evaluate the concepts learned by: 1️⃣ Neuro-Symbolic models (#NeSy) 2️⃣ Concept Bottleneck Models (#CBMs) 3️⃣ Black-box Neural Networks (NNs*) 4️⃣ Vision-Language Models (#VLMs*) * through post-hoc concept-based explanations (e.g., TCAV)
    Scarica
    103403
  10. Emanuele Marconato@ema_marconato
    Check out our latest paper @ #UAI2024. 🚀 We introduce BEARS 🐻, a method to detect wrongly learned concepts in #nesy models, due to Reasoning Shortcuts!
    Samuele Bortolotti @samubortolotti
    #NeuroSymbolic #AI 🤖 enforces constraints, but models can achieve high accuracy using wrong concepts. Can we spot when a model relies on flawed concepts and ensure #trustworthiness? Yes, with BEARS! 🐻 📢 Introducing our latest paper, accepted as a spotlight at UAI2024 !🎉📄
    11241.4K
  11. Samuele Bortolotti@samubortolotti
    #NeuroSymbolic #AI 🤖 enforces constraints, but models can achieve high accuracy using wrong concepts. Can we spot when a model relies on flawed concepts and ensure #trustworthiness? Yes, with BEARS! 🐻 📢 Introducing our latest paper, accepted as a spotlight at UAI2024 !🎉📄
    18386.6K
  14. Alexander Koller@alkoller
    Come do research with me, my fantastic colleagues, and some of the coolest PhD students I've ever met! #NLProc #nesy #neurosymbolic #AI #ML
    Neuroexplicit Models of Language, Vision, Action @neuroexplicit
    Now hiring: Twelve (!) PhD students to start in fall 2025, for research on combining neural and symbolic/interpretable models of language, vision, and action. Work with world-class advisors at @Saar_Uni, MPI Informatics, @mpi_sws, @CISPA, @DFKI. Details:
    12102.7K
  17. Tilman Hinnerichs@thinnerichs
    I presented our work "Towards a fully declarative neuro-symbolic language" at #NeSy 2024. Maybe this is the NeSy answer to generative AI. 👀 @EmilevanKrieken Check it out on Arxiv:
    Ernesto Jimenez-Ruiz @ejimenez_ruiz
    #NeSy 2024 Day 2
    15201.5K
  20. Eleonora Giunchiglia@e_giunchiglia
    Want to use your favourite #NeSy model but afraid of the reasoning shortcuts?🫣 Fear not💪🏻In our #NeurIPS2025 paper we show that you just need to equip your favourite NeSy model with prototypical networks and the reasoning shortcuts will be a problem of the past! w/ @GdiGiotto
    Scarica
    111202.1K
  21. Olivia Wang@oliviawpy2023
    Just by adding the DFA that constraints the reasoning to include the number at least once, it improved the performance. This is a good illustration of how simple rules can significantly improve the performance of LLMs. #Neurosymbolic #Nesy
    Scarica
    113159
  22. Daniel Dobriy (Q112275832)@semantic42
    🎤 Keynote at SymGenAI4Sci (SEMANTiCS 2025): Robert Tjarko Lange ( on “Toward Open‑Ended Agentic Discovery: LLMs as Artists, Scientists, and Evolutionary Engines” 📅 Sept 3, 13:00 CEST Vienna Details: #SEMANTiCS #NeSy #GenAI
    Scarica
    13161.8K
  23. Samuele Bortolotti@samubortolotti
    @samubortolotti
    Combining Deep Ensembles with diversification, BEARS optimizes all goals. Experiments show: - SoTA NeSy and CBM models are overconfident about their concepts. - BEARS increases uncertainty for concepts affected by RSs. - Performance stays high, predictions are more calibrated.
    Scarica
    1052K
  24. Samuele Bortolotti@samubortolotti
    @samubortolotti
    BEAR-calibrated #NeSy models make it easy to acquire concept supervision using uncertainty-based active learning, reducing the cost of RS mitigation. By asking users to annotate uncertain samples, BEARS quickly learns the right concepts while maintaining high label accuracy!
    Scarica
    105214
  25. Emanuele Marconato@ema_marconato
    Happy to present at #NeurIPS2025 If you want to know about our latest research on #Nesy AI and #interpretability come and meet us 😉
    Structured Machine Learning Group - Trento @SMLGroup_Trento
    Tomorrow (4/12) at poster session 4 (4.30pm - 7.30pm, poster 1011) @samubortolotti and @ema_marconato are presenting their work on identifiability, concept-based models and reasoning shortcuts;
    03202.6K
  27. Stefano Teso@looselycorrect
    📣 Do #NeSy #AI predictors consistent with prior #knowledge always learn high quality neural #concepts? Our new #NeurIPS2023 paper shows the answer is NO - due to #reasoning #shortcuts! 👋 trustworthiness! 😅 Wanna know more? Check this 🧵 by @ema_marconato!👇
    Emanuele Marconato @ema_marconato
    Using Neuro-Symbolic AI can boost the prediction accuracy... but are you learning the right concepts? 🤖 The answer is very tricky, but it is "no" in general 😱 Introducing our latest paper, accepted at #NeurIPS2023! With @tetraduzione, @andrea_whatever, and @looselycorrect.
    03181.3K
  29. Emanuele Marconato@ema_marconato
    🫕 We've been cooking something this year, and now is served at #NeurIPS2024! rsbench is a set of Neuro-symbolic tasks where learning concepts is hard because of Reasoning Shortcuts. Neural, Concept-based, NeSy, & Foundation models predict PERFECT labels, but WRONG concepts😱
    Samuele Bortolotti @samubortolotti
    📣 Does your model learn sensible #concepts? Test it with our #NeurIPS2024 benchmark! rsbench: A Neuro-Symbolic Benchmark Suite for Concept Quality and Reasoning Shortcuts @ema_marconato T Carraro @paolo_morettin @EmilevanKrieken @tetraduzione @looselycorrect @andrea_whatever
    34251.9K
  30. Emanuele Marconato@ema_marconato
    @ema_marconato
    People working in #concepts in #XAI and #NeSy, rsbench can help to understand further how models learn concepts: 🫰Easy to use, coming with default settings 👹Generative, can be personalized (more Reasonig Shortcuts) 📊Task/model-level metrics
    105158
  31. Pasquale Minervini@PMinervini
    Featuring our chapter on approximate complex query answering! #nesy #neuro #symbolic
    Scarica
    Pascal Hitzler @pascalhitzler
    The new edited book Compendium of Neurosymbolic Artificial Intelligence is now available - 30 chapters (700 pages), each conveying integrated content from several published papers of recent years.
    012368.6K
  32. Pascal Hitzler@pascalhitzler
    The new edited book Compendium of Neurosymbolic Artificial Intelligence is now available - 30 chapters (700 pages), each conveying integrated content from several published papers of recent years.
    1186423.1K
  33. Mostra altri post