Elamur Research⌕
☰
home / papers / 2026 / AI for Climate Finance: Agentic Retrieval and Multi-Step Reasoning for Early Warning System Investments

AI for Climate Finance: Agentic Retrieval and Multi-Step Reasoning for Early Warning System Investments

Paper · 2026 · Ario Saeid Vaghefi, Aymane Hachcham, Veronica Grasso et al.

Summary

Climate-finance reports from Multilateral Development Banks (MDBs) and climate funds rarely label spending on Early Warning Systems (EWS) consistently. They arrive as messy PDFs with complex tables, which makes it hard to track progress toward the UN Early Warnings for All (EW4All) goal of universal coverage by 2027.

The authors treat this as two joint tasks: assigning document snippets to one or more of the five EW4All pillars, and extracting per-pillar budget amounts linked to supporting evidence. Their "glass-box" agentic RAG pipeline works in stages:

  • it parses PDFs with Docling;
  • it enriches chunks with document-level summaries;
  • it retrieves passages by combining dense embeddings and BM25F lexical search through Reciprocal Rank Fusion;
  • an LLM agent issues iterative sub-queries and re-queries when evidence looks insufficient;
  • outputs are consolidated into a pillar-structured JSON.

On a manually annotated CREWS Fund corpus, the system beats four baselines. It also outperforms commercial black-box assistants on an MDB evidence set curated with WMO. The tool is publicly deployed and makes EWS funding flows more transparent and auditable.

Key points

  • On CREWS Fund data, the agent reaches 0.87 accuracy, 0.89 precision and 0.83 recall. This is roughly 8–14 pp above the best of four baselines: zero-shot, few-shot, fine-tuned transformer, and few-shot CoT+ICL. A budget prediction counts as correct only within ±5% of the gold amount.
  • On the WMO co-curated MDB set, median total-amount accuracy is about 0.78 for the agent, versus about 0.73 for Gemini 2.5 Flash and about 0.68 for OpenAI Assistants. Gemini does slightly better only on prose-heavy documents.
  • In ablations, removing context augmentation, hybrid retrieval or the self-healing loop each lowers performance, and top-k = 5 works best.
  • The deployed stack (ews-front.vercel.app) uses React/Vercel, FastAPI, Weaviate and the OpenAI API. It processes a 50-page report in under 3 minutes, compared with 2–3 hours of manual expert work.
  • The authors plan to release code, annotations and prompts. Low-confidence predictions (below 0.7) are flagged for expert review.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of ZurichZurich · Switzerland78 works · 12,260 citations

Also on the byline: Swiss Finance Institute · World Meteorological Organization.

 Fields, topics and concepts

📚 Shelves 7The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksInformation Systems · 1,437 worksEconomics, Econometrics and Finance · 17 worksFinance · 5 works

🧭 Topics 2Research subjects and languages the work is about.Reasoning · 1,426 worksInformation retrieval & search · 473 works

Cite

Cite as Ario Saeid Vaghefi, Aymane Hachcham, Veronica Grasso, Nakiete Msemo, Chiara Colesanti Senni and Markus Leippold. 2026. AI for Climate Finance: Agentic Retrieval and Multi-Step Reasoning for Early Warning System Investments. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 3: System Demonstrations).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-demo.34

@inproceedings{vaghefi2026ai,
  title = {{AI for Climate Finance: Agentic Retrieval and Multi-Step Reasoning for Early Warning System Investments}},
  author = {Ario Saeid Vaghefi, Aymane Hachcham, Veronica Grasso, Nakiete Msemo, Chiara Colesanti Senni and Markus Leippold},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-demo.34},
  url = {https://doi.org/10.18653/v1/2026.eacl-demo.34}
}

References

20 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Christoph Auer, Maksym Lysak, Ahmed Nassar, Michele Dolfi, Nikolaos Livathinos, Panos Vagenas, Cesar Berrospi Ramis, Matteo Omenetti, Fabian Lindlbauer, Kasper Dinkla, Lokesh Mishra, Yusik Kim, Shubham Gupta, Rafael Teixeira de Lima, Valery Weber, Lucas Morin, Ingmar Meijer, Viktor Kuropiatnyk, and… find
  2. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in Neural Information Processing Systems (NeurIPS), 33:1877–1901.
  3. Stephen Robertson and Hugo Zaragoza. 2009. The probabilistic relevance framework: Bm25 and beyond. Foundations and Trends in Information Retrieval, 3(4):333–389. find
  4. Andrew C Tupper and Carina J Fearnley. 2023. Mind the gaps in disaster early-warning systems—and fix them. Nature, 623:479. find
  5. Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, and Zheng Liu. 2024. Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. find
  6. Gordon V. Cormack, Charles L.A. Clarke, and Stephan Buettcher. 2009. Reciprocal rank fusion outperforms condorcet and individual rank learning methods. In Proceedings of the 32nd International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 758–759. ACM.
  7. Shizhe Diao, Pengcheng Wang, Yong Lin, Rui Pan, Xiang Liu, and Tong Zhang. 2024. Active prompting with chain-of-thought for large language models. find
  8. Taicheng Guo, Xiuying Chen, Yaqi Wang, Ruidi Chang, Shichao Pei, Nitesh V. Chawla, Olaf Wiest, and Xiangliang Zhang. 2024. Large language model based multi-agents: A survey of progress and challenges.
  9. Michael Günther, Isabelle Mohr, Daniel James Williams, Bo Wang, and Han Xiao. 2024. Late chunking: Contextual chunk embeddings using long-context embedding models. find
  10. Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. 2022. Large language models are zero-shot reasoners. Advances in neural information processing systems, 35:22199– 22213.
  11. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, et al. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing Systems, 33:9459–9474.
  12. Junwei Liu, Kaixin Wang, Yixuan Chen, Xin Peng, Zhenpeng Chen, Lingming Zhang, and Yiling Lou. 2024. Large language model-based agents for software engineering: A survey. find
  13. Zach Nussbaum, John X. Morris, Brandon Duderstadt, and Andriy Mulyar. 2024. Nomic embed: Training a reproducible long context text embedder.
  14. Gianluca Pescaroli, Sarah Dryhurst, and Georgios Marios Karagiannis. 2025. Bridging gaps in research and practice for early warning systems: new datasets for public response. Frontiers in Communication, 10:1451800. find
  15. Omar Velazquez, Gianluca Pescaroli, Gemma Cremen, and Carmine Galasso. 2020. A review of the technical and socio-organizational components of earthquake early warning systems. Frontiers in Earth Science, 8:533498. find
  16. Jie Wang, Alexandros Karatzoglou, Ioannis Arapakis, and Joemon M Jose. 2024. Reinforcement learning-based recommender systems with large language models for state reward and action modeling. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retr… find
  17. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2023. Self-consistency improves chain of thought reasoning in language models.
  18. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824–24837.
  19. Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, Enyu Zhou, Rui Zheng, Xiaoran Fan, Xiao Wang, Limao Xiong, Yuhao Zhou, Weiran Wang, Changhao Jiang, Yicheng Zou, Xiangyang Liu, Zhangyue Yin, Shihan Dou, Rongxiang Weng, Wensen Cheng, Qi Zhang,… — The Rise and Potential of Large Language Model Based Agents: A Survey
  20. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2023. ReAct: Synergizing reasoning and acting in language models. In International Conference on Learning Representations (ICLR).

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/vaghefi-2026-ai-climate-finance-agentic-retrieval/