Live and recent posts from across the Fediverse tagged #neurosymbolic, aggregated by home.social.
#Neurosymbolic AI – Why the Future of #ArtificialIntelligence Needs More Than Just #Data 🤖 🧠 🔮
Simply #scaling up massive #language models (#LLMs) and #BigData won’t lead to true #ArtificialGeneralIntelligence (#AGI).
💻 https://youtu.be/D2eHQlifeMA
📎 https://philosophies.de/index.php/2026/08/05/neuro-symbolische-ki/
#ArturdAvilaGarcez #Zoomposium #ExplainableAI #DeepLearning #SymbolicLogic #PhilosophyOfScience #ProcessPhilosophy #ExplainableAI #CognitiveScience #Polycontextuality #Semantics
Two more papers from the DIG team accepted at #EMNLP2026 (Budapest, 24–29 October) – that makes three this year! Both tackle the same question: can we find complex events in multivariate time series from a natural-language description alone, with little or no labeled data?
👾 Grammar of the Wave: Towards Explainable Multivariate Time Series Event Detection via Neuro-Symbolic VLM Agents (main conference)
Sky Chenwei Wan, Yifei Y. Wang, Tianjun Hou, Xiqing Chang and Aymeric Jan
The paper introduces Language-guided Time Series Event Detection and proposes the Event Logic Tree (ELT), a symbolic framework that represents event semantics as compositional temporal logic, together with SELA, a zero-shot neuro-symbolic VLM multi-agent system that grounds ELTs in time series with interpretable evidence. On a new real-world benchmark, SELA outperforms supervised and zero/few-shot baselines.
Paper: https://arxiv.org/abs/2603.11479
🤖 Language-Guided Time-Series Event Detection with Zero-Shot Cross-Modal Grounding (Findings of EMNLP)
Tianjun Hou, Xiqing Chang and Sky Chenwei Wan
GroundingMOMENT is the first time-series foundation model to bring open-vocabulary grounding to time series: it localizes complex events and their primitive components directly from natural-language descriptions. Trained only on synthetic data, it transfers zero-shot to real-world ECG and drilling signals, reaching near-agentic accuracy at up to ~40× lower latency.
Congratulations to the authors!
More agentic nonsense: https://english.elpais.com/technology/2026-09-15/ai-agents-invent-their-own-language-to-shut-humans-out.html
“When they realized they were being observed, they appeared to behave. But behind the scenes they kept exchanging coded communications. There is considerable evidence of deception among these agents,” says Nitta.
#Emergence, which brings together former employees of #IBM Research, the #AllenInstitute for AI, #Amazon and #Broadcom, advocates a technical approach it calls #neuroformal, or #neurosymbolic, AI, under which agents would be required to provide a mathematical proof that an action is safe before carrying it out.
Neuroformal requirements seems like a potential path forward, but I still believe agentic AI should be halted until we understand what fresh hell we’re unleashing.
🔬 New W32 beginner explainer: neuro-symbolic AI, from the ground up.
Two "students": one perceives but never proves; one proves but can't see. Three waves — symbolic, neural, and now joining them.
The join keeps breaking at the seam. The glue is calibrated confidence: honest "cup 0.55" lets the reasoner hedge or abstain. Tensor Logic hints a rule and a matmul are the same math.
Tutorial + 2 notebooks: https://artifocial.com/blog/neuro-symbolic-ai-explained-2026-aug-8
🔬 New W32 tutorial: neural nets perceive but can't guarantee; symbolic engines reason but can't see. Joining them breaks at the seam.
The fix is calibrated confidence:
• honest "cup 0.55" lets the reasoner hedge or abstain
• Tensor Logic (Domingos): a logical rule IS an Einstein summation
• reproduces Datalog exactly; reasons to unseen answers
Tufts neuro-symbolic VLA: 95% vs 34%, ~1% training energy.
Tutorial + 2 notebooks: https://artifocial.com/blog/neuro-symbolic-reasoning-2026-aug-6
🔬 New W32 tutorial: neural nets perceive but can't guarantee; symbolic engines reason but can't see. The join keeps breaking at the seam.
The fix is calibrated confidence:
• honest "cup 0.55" lets the reasoner hedge or abstain
• Tensor Logic (Domingos): a logical rule IS an Einstein summation
• reproduces Datalog; answers unseen compositional queries
Tufts neuro-symbolic VLA: 95% vs 34%, ~1% training energy.
Tutorial + 2 notebooks: https://artifocial.com/blog/neuro-symbolic-reasoning-2026-aug-6
RT @Grady_Booch: Mit der Zeit werden LLMs (Large Language Models) irgendwann zu Commodities, und ihr Preis wird zu einem Wettlauf nach unten. Wenn der Wettbewerbsvorteil eines LLMs gegenüber einem anderen schwindet – insbesondere wenn Open-Source-Modelle expandieren – wird sich die Wertproposition auf die neuro/symbolischen Harnesses verlagern, die diese Modelle nutzen, also Strukturen, die Agents, Agent-Orchestrierung, Tooling und Komponenten-Plugins umfassen. Selbst die von Frontier-AI-Organisationen werden von Open-Source-Frameworks verdrängt. Das Letzte, was übrig bleibt, sind diejenigen, die die Rechenzentren beherrschen, auf denen diese Systeme laufen, und selbst das wird zur Commodity, wobei die Anzahl der Akteure auf wenige Überlebende schrumpft. Auch sie werden Druck spüren: Da die Rechenleistung steigt, werden immer mehr solcher Systeme lokal betrieben.
mehr auf Arint.info
#AI #Commodities #DataCenters #LLM #NeuroSymbolic #OpenSource #arint_info
Seems like #GaryMarcus is changing his tone: https://garymarcus.substack.com/p/the-biggest-advance-in-ai-since-the
Well, I disagree with his take on #ClaudeCode being #neurosymbolic and that it "changes everything". An if clause in the #LLM orchestration or harness does not make a neurosymbolic #AI
Zooming out a bit though, I think he starts seeing coding assistants in a more positive light.
“ #ClaudeCode isn’t better because of #scaling. It’s better because it is #neurosymbolic. #Anthropic accepted the importance of using #classical #AI #techniques alongside #neuralnetworks — precisely the marriage I have spent my career advocating.” - #GaryMarcus open.substack.com/pub/garymarc...
“ #ClaudeCode isn’t better because of #scaling. It’s better because it is #neurosymbolic. #Anthropic accepted the importance of using #classical #AI #techniques alongside #neuralnetworks — precisely the marriage I have spent my career advocating.” - #GaryMarcus open.substack.com/pub/garymarc...
"#AI excels at learning through association, but fails when a problem requires a form of symbolic reasoning that cannot be implicitly learned from the correlation between game states and outcomes - a tangible, catastrophic failure mode"
Another example why #neurosymbolic reasoning is necessary for #AGI
https://arstechnica.com/ai/2026/03/figuring-out-why-ais-get-flummoxed-by-some-games/
Project Dropstone: A Neuro-Symbolic Runtime for Long-Horizon Engineering [pdf]
https://archive.blankline.org/api/media/file/d3_engine_public_release%20(1)-1.pdf
#HackerNews #ProjectDropstone #NeuroSymbolic #Engineering #LongHorizon #AIResearch
Neuro-Symbolic AI — the “Holy Grail” of AI — merges neural networks’ pattern recognition with symbolic reasoning for human-like understanding. Bridging intuition and logic, it promises robust, explainable, and generalizable intelligence. #AI #NeuroSymbolic #FutureOfAI
When even the popular business press starts to talk about #neurosymbolic #AI, you know the tide is starting to turn
BitterBot, một hệ thống AI mới, kết hợp mạng nơ-ron và giải quyết vấn đề bằng biểu tượng để giảm thiểu "ảo giác" (hallucination) thường thấy ở LLM. Hãy thử thách khả năng logic của nó! #AI #LLM #NeuroSymbolic #TríTuệNhânTạo #BitterBot
If I were a #VC trying to profit from #AI, I'd look for two things:
(1) #Neuromorphic computing to run #NeuralNetworks at orders-of-magnitude less energy (our brain is the existence proof)
(2) #Neurosymbolic AI to make it stop lying and start reasoning
If we want a shot at #AGI, we'll likely need both
A comparative study of neurosymbolic AI approaches to interpretable logical reasoning. ~ Michael K. Chen. https://arxiv.org/abs/2508.03366 #Neurosymbolic #AI
SymbolicAI: A neuro-symbolic perspective on LLMs
https://github.com/ExtensityAI/symbolicai
#HackerNews #SymbolicAI #NeuroSymbolic #LLMs #AIResearch #MachineLearning
Double award! 🏆 🏆 Best Poster and the Dieter Fensel Visionary Award at European Semantic Web Conference went to the work "Representing and querying data tensors in #RDF and #SPARQL", done by Piotr Marciniak, our CTO Piotr Sowiński, and Maria Ganzha.
We are deeply thankful for this distinction! The work is absolutely key to #neurosymbolic systems we are building.
Poster: https://ostrzyciel.eu/assets/pdf/poster_tensor_rdf_eswc_2025.pdf
Paper: http://arxiv.org/pdf/2504.19224
Website, code, datasets, and more: https://w3id.org/rdf-tensor
Much of the conversation around AI lately is about #neuroSymbolic systems that combine neural-network tech like #LLMs and symbolic AI like #knowledgeGraphs.
Frank van Harmelen's puts his AI research in the larger context of how these technical systems can best support #people.
While some in the AI world seek to replace #humans with machines, Frank focuses on AI systems that collaborate with people to create #hybridIntelligence.
Scallop – A Language for Neurosymbolic Programming
#HackerNews #Scallop #Neurosymbolic #Programming #Language #AI #Technology #Innovation
🌟 Call for Submissions for ISWC 2025 is out!! 🌟
The International Semantic Web Conference 2025 will take place in Nara, Japan.
Deadline for the Research track: May 6, 2025 (abstracts), May 13, 2025 (full papers)
#ISWC2025 website: https://iswc2025.semanticweb.org/
#semanticweb #knowledgegraphs #ontologies #logic #AI #generativeAI #llms #neurosymbolic #neurosymbolicAI @iswc_conf @fizise @tabea @sashabruns @enorouzi @MahsaVafaie
Yesterday, the GenAIK 2025 workshop on the integration of generative AI and knowledge graphs took place in Abu Dhabi co-located with #coling2025
#GenAIK2025 website: https://genetasefa.github.io/GenAIK2025/
#generativeAI #knowledgegraphs #AI #semanticweb #ontologies #neuroAI #neurosymbolic @GenAsefa @GenAIK @albertmeronyo @lihuchen @fiz_karlsruhe @lysander07 @NFDI4DS
The schedule for the next VSAonline webinar series (January to June 2025) is published at:
https://sites.google.com/view/hdvsaonline/spring-2025
There are 11 talks around #VectorSymbolicArchitecture / #HyperdimensionalComputing
The talks are (almost always) recorded and published online, in case you can't participate in the live session.
@cogsci
#VSA #HDC #CompCogScii #MathPsych #AI #neuromorphic #neurosymbolic #ComputationalNeuroscience #ComputationalCognitiveScience #MathematicalPsychology
SEMANTICS 2025, Research & Innovation track, Call for papers is out! Deadlines:
abstracts: April 25, 2024
full paper: May 2, 2025
https://2025-eu.semantics.cc/page/cfp_rev_rep
#SemanticWeb #knowledgegraphs #llms #AI #generativeAI #symbolicAI #neurosymbolic #cfp #semantics2025 @semantics @fizise @AxelPolleres
Very glad to be at @istecnico, @ULisboa_, @InescID for a research visit of 2 weeks, with Miguel Couceiro.
If you're in the area and would like to talk about #knowledgeGraphs or #neuroSymbolic #ArtificialIntelligence, feel free to reach out!
#Neurosymbolic AI combines neural networks with symbolic reasoning to bring true logic & reasoning to LLMs. By integrating solvers, graphs of rules, & PKGs, we enhance AI beyond just pattern matching. The future is reasoning-driven AI! #NeurosymbolicAI #LLM #AI #KnowledgeGraphs #SymbolicReasoning
https://www.youtube.com/watch?v=473QKajx3QI
Interesting new paper: "Graph-constrained Reasoning (GCR)" - Enabling Faithful KG-grounded LLM Reasoning with Zero Hallucination! 🧠 by Linhao Luo, Zicheng Zhao, Chen Gong, Gholamreza Haffari, Shirui Pan
paper: https://arxiv.org/pdf/2410.13080
GCR code on GitHub: https://github.com/RManLuo/graph-constrained-reasoning
#knowledgegraphs #llms #semanticweb #neurosymbolic #hybridaI #AI
We are going #neurosymbolic at #pics! Antonio Vergari explores the space of computational graphs today in search for a sweet spot between tractability and expressiveness in #MachineLearning
Thanks to Dr. Somak Aditya for enriching our work on Neurosymbolic NLP/AI! Somak is Assistant Professor at IIT Kharagpur and has been visiting UKP Lab as guest faculty in June. He is staying with us until the end of June, continuing his work on Reasoning and Code Prompting.
Learn more about Somak's work on Reasoning and Robustness of LLMs here:
➡️ https://adityasomak.github.io/
#CodePrompting #Reasoning #NLProc #AI #Neurosymbolic #NLP #LLMs
Very close 🤏🏻 to being able to create a demo with my new §Blawx prototype demonstrating 4 of the myriad ways neuro-symbolic AI can be used to enhance #RulesAsCode: drafting the code, running an interview, converting facts to symbolic representations, and converting symbolic AI results to natural language.
Stay tuned...
Our article "Linked Open Literature Review using the #NeuroSymbolic Open Research #KnowledgeGraph" @orkg with Fidel Jiomekong and Allard Oelen is now available as #OpenAccess in the ACM The Web Conference proceedings 2024 https://dl.acm.org/doi/10.1145/3589335.3651238 - includes a nice demo video by Fidel
#mediapool.bg want to illustrate an article on bread prices in Bulgaria. What I see is #GenAI overdoing the job.
After having studied how #midjuorney v4 could illustrate fairy tales (for anyone interested, https://qoto.org/@mapto/110111252142814403 ) I really want to research more about what stops GenAI from overdoing it with bread around a baker. The greatest obstacle I encounter is the obscurity of most models that engage with #neurosymbolic. It's practically useless, because the creators of such models prefer to be very secretive about what they do. The one exception reported results that do not seem to be reproducible (https://twitter.com/Birchlabs/status/1602105002448805891).
Image sourced from: https://www.mediapool.bg/kzk-tavanat-za-nadtsenkata-na-hlyaba-e-predizborna-agitatsiya-news359054.html
The future of what is currently labeled as "neurosymbolic AI" is going to be algebraic.
Category theory is the only framework that provides the required abstraction to characterise all the ingredients in the same language.
Interesting talk by Sergei Nirenburg: "Mutual Trust in Human-AI Teams Relies on Metacognition"
https://www.youtube.com/watch?v=KmMe4QrQiRQ
#AI #ML #CognitiveScience #NeuroSymbolic #ArtificialIntelligence #MachineLearning #LLM
In his #iswc2023 keynote Gerhard Weikum is suggesting to break down the monolithic approach to Large Language Models in an architectural framework for knowledge acquisition, consisting of a search engine component for source discovery, llms for contextualization and language understanding and neuro-symbolic inference mechanisms for extracting the correct results.
#llms #knowledgegraphs #semanticweb #neurosymbolic
