Seed Papers for Survey

We need to come up with a list of papers which can be used as seed or root papers for getting the list of papers to be surveyed in "AI for Mathematics".

1. Survey of Deep Learning for Theorem Proving
   • Github repo link
   • GS shows 40+ citations
2. Advancing mathematics by guiding human intuition with AI
   • GS shows around 700+ citations
   • Harvard ADS shows around 140+ citations
3. Deep learning for symbolic mathematics
   • The [connectedPapers link] shows around TODO
   • GS shows around 640+ citations
   • Harvard ADS shows around 160+ citations
4. LeanDojo: Theorem Proving with Retrieval-Augmented Language Models
   • The connectedPapers link shows around 300+ citations
   • GS shows around 390+ citations
   • The Harvard link above shows 78 citations
5. Lean Copilot: Large Language Models as Copilots for Theorem Proving in Lean
   • Harvard ADS shows around 50 citations
   • GS shows around 9 citations
6. The Calabi-Yau Landscape: from Geometry, to Physics, to Machine-Learning
   • GS shows around 90+ citations
   • Harvard ADS shows around 60+ citations
7. Minif2f: A Cross-system Benchmark for Formal Olympiad-level Mathematics
   • GS shows around 280+ citations
   • Harvard ADS lists around 120+ citations
8. miniF2F-Lean Revisited: Reviewing Limitations and Charting a Path Forward
   • GS shows no citations
   • Harvard ADS shows no citations
9. Neural Theorem Proving: Generating and Structuring Proofs for Formal Verification
   • GS shows only one citation
10. APOLLO: Automated LLM and Lean Collaboration for Advanced Formal Reasoning
    • ConnectedPapers link shows around 40+ citations
    • GS shows only 7 citations
    • Harvard ADS shows around 7 citations
11. Machine-learning mathematical structures
    • GS shows around 60+ citations
    • Harvard ADS shows around 20+ citations
12. Aristotle: IMO-level Automated Theorem Proving
    • Harvard ADS shows around 1 citation
13. BliStrTune: Hierarchical Invention of Theorem Proving Strategies
    • GS shows 30+ citations
14. An Extensible Theorem Proving Frontend
    • ConnectedPapers shows 3 citations
    • GS shows only 10 citations
15. Reviving DSP for Advanced Theorem Proving in the Era of Reasoning Models
    • Harvard ADS shows around 6 citations
16. REFACTOR: Learning to Extract Theorems from Proofs
    • GS shows around 10 citations
    • Harvard ADS shows around 3 citations

Youtube Videos

1. Interview: Building an AI Mathematician
2. Interview w/ Christian Szegedy: Autoformalization and Verifiable Superintelligence
3. AI for Math Resources - GDoc
4. TODO - Add others

TODOs

• Convert top-level links to arxiv abstracts for the ones which are not.
• Create consolidated list of references from the root papers identified above
  • Create script to scrape the GS citations from browser
  • Create script to scrape the Harvard ADS citations list
  • One can download the bib file of citations from connectedpapers, so no need for script.
• From this repo, select the additional root papers to consider.
• Come up with a category for classifying the papers.
  • The groups in the above link & the survey paper repo can be used.