I think their general purpose reasoning model is an LLM. The contrast is with domain specific things like AlphaGeometry. The prompts are natural language.
May 21, 2026 at 12:42 PM
I think their general purpose reasoning model is an LLM. The contrast is with domain specific things like AlphaGeometry. The prompts are natural language.
Providence/greater-Boston-area people: check out this talk on March 18 (register at ICERM site) by Junehyuk Jang, a Brown professor and member of the DeepMind team. A success of neuro-symbolic reasoning!
March 10, 2025 at 1:53 PM
Providence/greater-Boston-area people: check out this talk on March 18 (register at ICERM site) by Junehyuk Jang, a Brown professor and member of the DeepMind team. A success of neuro-symbolic reasoning!
Spent last week @ias with DeepMind math teams. AlphaProof, AlphaGeometry, PDEs, and PAIRs teams were there with leadership. Impressed by the collaboration. Huge, open technical problems to be solved. Exciting times for mathematics & AI!
www.ias.edu/math/events/...
www.ias.edu/math/events/...
May 6, 2025 at 12:03 PM
Spent last week @ias with DeepMind math teams. AlphaProof, AlphaGeometry, PDEs, and PAIRs teams were there with leadership. Impressed by the collaboration. Huge, open technical problems to be solved. Exciting times for mathematics & AI!
www.ias.edu/math/events/...
www.ias.edu/math/events/...
This is a bit misleading. The first link is to AlphaProof/AlphaGeometry 2, which are not general LLMs like the ones shown in the image—they are neurosymbolic models trained specifically for Olympiad-type problems.
I don't think it comes as a surprise that general-purpose LLMs still struggle.
I don't think it comes as a surprise that general-purpose LLMs still struggle.
April 2, 2025 at 9:46 AM
This is a bit misleading. The first link is to AlphaProof/AlphaGeometry 2, which are not general LLMs like the ones shown in the image—they are neurosymbolic models trained specifically for Olympiad-type problems.
I don't think it comes as a surprise that general-purpose LLMs still struggle.
I don't think it comes as a surprise that general-purpose LLMs still struggle.
before alphageometry (etc) there was alphafold; nice essay below on the problems the latter shares with the former, much obscured by the hype
New sanctioned (t)rant about AI just dropped. On the limits of Alphafold and my worries about the disaster its use will cause for the literature and early-stage academic drug discovery.
Look, they commissioned it. And by commissioned I don't mean I got paid, just invited. Awesome process btw.
Look, they commissioned it. And by commissioned I don't mean I got paid, just invited. Awesome process btw.
The AI Chemist: To be trustworthy, LLMs need to show their work
Good scientists reveal how they do their experiments and report their results; so should any machine-driven research
cen.acs.org
August 1, 2026 at 11:09 AM
before alphageometry (etc) there was alphafold; nice essay below on the problems the latter shares with the former, much obscured by the hype
Davis and Marcus dive deep in AlphaGeometry 2, deflating hype but also pointing to some genuine positives.
garymarcus.substack.com/p/alphageome...
garymarcus.substack.com/p/alphageome...
AlphaGeometry2: Impressive accomplishment, but still a long path ahead
What GoogleDeepMind’s latest does and doesn’t show, and what we like about it
garymarcus.substack.com
February 17, 2025 at 5:27 PM
Davis and Marcus dive deep in AlphaGeometry 2, deflating hype but also pointing to some genuine positives.
garymarcus.substack.com/p/alphageome...
garymarcus.substack.com/p/alphageome...
AI just leveled up in math—and not just any math, but the legendary International Mathematical Olympiad (IMO)! 🤖✨
DeepMind’s AlphaProof and AlphaGeometry 2 tackled this year’s IMO problems, solving 4 out of 6—matching the silver medalist tier for the first time ever.
DeepMind’s AlphaProof and AlphaGeometry 2 tackled this year’s IMO problems, solving 4 out of 6—matching the silver medalist tier for the first time ever.
July 22, 2025 at 4:32 PM
AI just leveled up in math—and not just any math, but the legendary International Mathematical Olympiad (IMO)! 🤖✨
DeepMind’s AlphaProof and AlphaGeometry 2 tackled this year’s IMO problems, solving 4 out of 6—matching the silver medalist tier for the first time ever.
DeepMind’s AlphaProof and AlphaGeometry 2 tackled this year’s IMO problems, solving 4 out of 6—matching the silver medalist tier for the first time ever.
"AlphaGeometry" Matematik Olimpiyatı düzeyinde problemleri çözüyor. www.nature.com/articles/d41...
Google DeepMind karmaşık geometri problemlerini çözebilen bir yapay zeka sistemi yaratmış. Uzmanlara göre bu, insan benzeri düşünme becerilerine sahip makinelere doğru atılmış önemli bir adımmış.
www.technologyreview.com/2024/01/17/1...
www.technologyreview.com/2024/01/17/1...
January 18, 2024 at 1:23 PM
"AlphaGeometry" Matematik Olimpiyatı düzeyinde problemleri çözüyor. www.nature.com/articles/d41...
𝐃𝐞𝐞𝐩𝐌𝐢𝐧𝐝𝐬 𝐀𝐥𝐩𝐡𝐚𝐆𝐞𝐨𝐦𝐞𝐭𝐫𝐲 𝐬𝐜𝐡𝐥ä𝐠𝐭 𝐌𝐚𝐭𝐡𝐞-𝐎𝐥𝐲𝐦𝐩𝐢𝐚𝐝𝐞𝐧-𝐆𝐨𝐥𝐝𝐦𝐞𝐝𝐚𝐢𝐥𝐥𝐞𝐧𝐠𝐞𝐰𝐢𝐧𝐧𝐞𝐫
KI löst komplexe Geometrieprobleme
Übertrifft menschliche Experten
Durchbruch in der Mathematik
#ai #ki #artificialintelligence #DeepMind #AlphaGeometry #Mathematik
kinews24.de/deepmind-alp...
KI löst komplexe Geometrieprobleme
Übertrifft menschliche Experten
Durchbruch in der Mathematik
#ai #ki #artificialintelligence #DeepMind #AlphaGeometry #Mathematik
kinews24.de/deepmind-alp...
DeepMind AlphaGeometry übertrifft Goldmedaillengewinner der Internationalen Mathematik-Olympiade
DeepMind AlphaGeometry hat einen bemerkenswerten Durchbruch erzielt. Die KI hat in einem Test, der für Goldmedaillengewinner der Internationalen Mathematik-Olympiade (IMO) entwickelt wurde, besser abg...
kinews24.de
February 9, 2025 at 9:43 AM
𝐃𝐞𝐞𝐩𝐌𝐢𝐧𝐝𝐬 𝐀𝐥𝐩𝐡𝐚𝐆𝐞𝐨𝐦𝐞𝐭𝐫𝐲 𝐬𝐜𝐡𝐥ä𝐠𝐭 𝐌𝐚𝐭𝐡𝐞-𝐎𝐥𝐲𝐦𝐩𝐢𝐚𝐝𝐞𝐧-𝐆𝐨𝐥𝐝𝐦𝐞𝐝𝐚𝐢𝐥𝐥𝐞𝐧𝐠𝐞𝐰𝐢𝐧𝐧𝐞𝐫
KI löst komplexe Geometrieprobleme
Übertrifft menschliche Experten
Durchbruch in der Mathematik
#ai #ki #artificialintelligence #DeepMind #AlphaGeometry #Mathematik
kinews24.de/deepmind-alp...
KI löst komplexe Geometrieprobleme
Übertrifft menschliche Experten
Durchbruch in der Mathematik
#ai #ki #artificialintelligence #DeepMind #AlphaGeometry #Mathematik
kinews24.de/deepmind-alp...
Trois petites histoires autour de thème de l'automatisation du raisonnement géométrique :
* L'axiome de Pasch et les limites de la géométrie d'Euclide
* Les preuves informatisées avec l'exemple de GeoCoq
* AlphaGeometry et le raisonnement « neuro-symbolique »
lipsum.dev/2024-07-1-me...
* L'axiome de Pasch et les limites de la géométrie d'Euclide
* Les preuves informatisées avec l'exemple de GeoCoq
* AlphaGeometry et le raisonnement « neuro-symbolique »
lipsum.dev/2024-07-1-me...
Quelques aspects de la mécanisation de la géométrie
En début d'année 2024, des chercheurs de DeepMind annoncent AlphaGeometry :
un logiciel capable de résoudre des problèmes de géométrie posés aux Olympiades…
lipsum.dev
January 1, 2025 at 9:22 PM
Trois petites histoires autour de thème de l'automatisation du raisonnement géométrique :
* L'axiome de Pasch et les limites de la géométrie d'Euclide
* Les preuves informatisées avec l'exemple de GeoCoq
* AlphaGeometry et le raisonnement « neuro-symbolique »
lipsum.dev/2024-07-1-me...
* L'axiome de Pasch et les limites de la géométrie d'Euclide
* Les preuves informatisées avec l'exemple de GeoCoq
* AlphaGeometry et le raisonnement « neuro-symbolique »
lipsum.dev/2024-07-1-me...
AI has fought over two 60-year-old traditions: symbolic (rules) and neural (pattern-matching). Neither won — the best systems use both. AlphaGeometry solved 25/30 Olympiad geometry problems this way.
sgopalanbtg.substack.com/p/explainer-...
#AI #Neurosymbolic
sgopalanbtg.substack.com/p/explainer-...
#AI #Neurosymbolic
Explainer: Neural vs. Symbolic AI — The Two Traditions Behind Everything
Sriram Gopalan's thoughts on AI
sgopalanbtg.substack.com
September 23, 2026 at 8:14 PM
AI has fought over two 60-year-old traditions: symbolic (rules) and neural (pattern-matching). Neither won — the best systems use both. AlphaGeometry solved 25/30 Olympiad geometry problems this way.
sgopalanbtg.substack.com/p/explainer-...
#AI #Neurosymbolic
sgopalanbtg.substack.com/p/explainer-...
#AI #Neurosymbolic
"This achievement is a significant advance over last year. At IMO 2024, AlphaGeometry and AlphaProof required experts to first translate problems from natural language into domain-specific languages, such as Lean, and vice-versa for the proofs. It also took two to three days of computation. [...]
July 21, 2025 at 4:59 PM
"This achievement is a significant advance over last year. At IMO 2024, AlphaGeometry and AlphaProof required experts to first translate problems from natural language into domain-specific languages, such as Lean, and vice-versa for the proofs. It also took two to three days of computation. [...]
let’s appreciate that this one paragraph has more real info than the 30+ X threads that OpenAI shoved out there on Saturday
July 22, 2025 at 12:34 AM
let’s appreciate that this one paragraph has more real info than the 30+ X threads that OpenAI shoved out there on Saturday
Machine learning and mathematics (The twin catalysts poised to accelerate twenty-first century mathematics: formalisation and machine learning). ~ Harald Carlens. joltml.com/ml-mathemati... #Math #ITP #Lean4 #MachineLearning #LLMs #AIMO #AlphaProof #AlphaGeometry
November 29, 2024 at 8:18 AM
Machine learning and mathematics (The twin catalysts poised to accelerate twenty-first century mathematics: formalisation and machine learning). ~ Harald Carlens. joltml.com/ml-mathemati... #Math #ITP #Lean4 #MachineLearning #LLMs #AIMO #AlphaProof #AlphaGeometry
Despite the disclaimer, I do think neurosymbolic AI - as DeepMind's AlphaGeometry is starting to show - is huge with potential.
February 1, 2024 at 12:06 AM
Despite the disclaimer, I do think neurosymbolic AI - as DeepMind's AlphaGeometry is starting to show - is huge with potential.
Google DeepMindが数学オリンピックレベルの幾何学問題を解けるAI「AlphaGeometry」を発表、人間の金メダリストに近い性能を発揮
https://gigazine.net/news/20240118-google-deepmind-alphagemetry/
https://gigazine.net/news/20240118-google-deepmind-alphagemetry/
Google DeepMindが数学オリンピックレベルの幾何学問題を解けるAI「AlphaGeometry」を発表、人間の金メダリストに近い性能を発揮
Google DeepMindが、国際数学オリンピックレベルの複雑な幾何学問題を解決できるAI「AlphaGeometry」を発表しました。AlphaGeometryは、実際に国際数学オリンピックで出題された幾何学問題30問を制限時間以内に25問解いたとのことです。
gigazine.net
January 18, 2024 at 2:09 AM
Google DeepMindが数学オリンピックレベルの幾何学問題を解けるAI「AlphaGeometry」を発表、人間の金メダリストに近い性能を発揮
https://gigazine.net/news/20240118-google-deepmind-alphagemetry/
https://gigazine.net/news/20240118-google-deepmind-alphagemetry/
Oui je suis assez d'accord avec vous, la puissance de AlphaGeometry reside surtout à mon sens dans une généralisation de la méthode "Deductive Database" avec un pivot de Gauss pour automatiser certains calculs. Pas clair que le LLM ne puisse pas être remplacé par une heuristique.
August 28, 2025 at 12:25 PM
Oui je suis assez d'accord avec vous, la puissance de AlphaGeometry reside surtout à mon sens dans une généralisation de la méthode "Deductive Database" avec un pivot de Gauss pour automatiser certains calculs. Pas clair que le LLM ne puisse pas être remplacé par une heuristique.
Readings shared November 29, 2024. jaalonso.github.io/vestigium/po... #ITP #Lean4 #IsabelleHOL #Logic #Math #Calculemus #MachineLearning #LLMs #AIMO #AlphaProof #AlphaGeometry #Haskell #FunctionalProgramming #Programming #FormalVerification #Dafny
November 30, 2024 at 7:11 AM
Readings shared November 29, 2024. jaalonso.github.io/vestigium/po... #ITP #Lean4 #IsabelleHOL #Logic #Math #Calculemus #MachineLearning #LLMs #AIMO #AlphaProof #AlphaGeometry #Haskell #FunctionalProgramming #Programming #FormalVerification #Dafny
January 17, 2024 at 5:15 PM
AlphaGo, AlphaGeometry, AlphaFold, etc... all RL based
March 10, 2025 at 8:00 PM
AlphaGo, AlphaGeometry, AlphaFold, etc... all RL based
Voir aussi cette discussion bsky.app/profile/jnar... concernant l'importance du LLM dans AlphaGeometry.
Oui je suis assez d'accord avec vous, la puissance de AlphaGeometry reside surtout à mon sens dans une généralisation de la méthode "Deductive Database" avec un pivot de Gauss pour automatiser certains calculs. Pas clair que le LLM ne puisse pas être remplacé par une heuristique.
October 21, 2025 at 8:58 PM
Voir aussi cette discussion bsky.app/profile/jnar... concernant l'importance du LLM dans AlphaGeometry.
Google DeepMind, matematiksel akıl yürütme konusunda uzmanlaşmış yapay zeka modeli AlphaProof'u ve geometri odaklı modelin güncellenmiş bir versiyonu olan AlphaGeometry 2'yi tanıtmış.
Google DeepMind'ın yapay zeka sistemleri artık karmaşık matematik problemlerini çözebiliyor.
Google DeepMind'ın yapay zeka sistemleri artık karmaşık matematik problemlerini çözebiliyor.
Google DeepMind’s AI systems can now solve complex math problems
AlphaProof and AlphaGeometry are steps toward building systems that can reason, which could unlock exciting new capabilities.
www.technologyreview.com
July 25, 2024 at 4:14 PM
Google DeepMind, matematiksel akıl yürütme konusunda uzmanlaşmış yapay zeka modeli AlphaProof'u ve geometri odaklı modelin güncellenmiş bir versiyonu olan AlphaGeometry 2'yi tanıtmış.
Google DeepMind'ın yapay zeka sistemleri artık karmaşık matematik problemlerini çözebiliyor.
Google DeepMind'ın yapay zeka sistemleri artık karmaşık matematik problemlerini çözebiliyor.
These kinds of neurosymbolic systems – combining symbolic DSLs, search trees, and LLMs – are some of the more interesting things currently going on in AI imo.
Gold-medalist Performance in Solving Olympiad Geometry with AlphaGeometry2
We present AlphaGeometry2, a significantly improved version of AlphaGeometry introduced in Trinh et al. (2024), which has now surpassed an average gold medalist in solving Olympiad geometry problems. ...
arxiv.org
February 7, 2025 at 8:05 PM
These kinds of neurosymbolic systems – combining symbolic DSLs, search trees, and LLMs – are some of the more interesting things currently going on in AI imo.