Agent-Guided Discovery of Exchange-Correlation Functionals with Physics-based Validation
Source check: Source checked. Not independent scientific confirmation.
Finding: Full arXiv Methods/Results describe LLM-agent-guided symbolic search of exchange-correlation functionals with self-consistent physics filters and held-out/external validation; the selected O159 functional retains an 11.7% held-out error improvement over ωB97M-V and comparable external GSCDB137 performance. The result is a computational chemistry functional candidate, not experimental validation.
Technology: LLM-agent-guided symbolic search of density-functional exchange-correlation forms.
Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.
- AI contribution
- Frontier LLM agents propose exchange-correlation functional forms evaluated through a GPU fitting and self-consistent physics workflow.
- Test and result
- Agent-derived O159 was evaluated self-consistently on held-out MGCDB84 and on 36 GSCDB137 datasets; it improved held-out error versus ωB97M-V, while remaining computational rather than experimental validation.
- Limitation / still unknown
- External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://arxiv.org/html/2611.13329
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.