MatSemNet uses LLMs to mine 700+ papers, modeling reaction pathways as sequences for catalyst discovery
bravo_abad · x · 2026-09-20
Liu and coauthors introduce MatSemNet, which uses an LLM to extract information from over 700 catalysis papers. Key ideas:
- Preserve context, don't flatten: the same material can behave very differently depending on electrolyte, pH, voltage and pathway, so the method keeps textual descriptions, numerical conditions and literature-reported reaction pathways as distinct information types rather than collapsing them into hand-built descriptors.
- Pathways as sequences: a dedicated LSTM learns ordering patterns of intermediates (e.g. NO₃ → NO₂), combined with experimental conditions.
The authors argue this reaction-context-aware representation improves catalyst discovery for tasks like nitrate-to-ammonia conversion.
More from Research
- LLM attacks 800+ graph theory conjectures, yields 30+ full proofs — marc_lelarge · 2026-09-20
- AI agents force academia to rethink credit assignment, junior researchers warn — YiMaTweets · 2026-09-20
- ZGCM-1 Chinese dataset trends on Hugging Face with 1B-10B entries — zgcagi · 2026-09-20
- Gary Marcus disputes LLM 'pain direction' paper: language clusters don't mean suffering — anilkseth · 2026-09-20
- Smart Zoi runs 50 AI NPCs in real time with a fine-tuned 1B SLM, LIFT hits 100+ citations — Kangwook_Lee · 2026-09-20
- LlamaIndex benchmarks 100+ models on document parsing with ParseBench — solyarisoftware · 2026-09-20