Elamur Research⌕
☰
home / papers / 2026 / IF-GEO: Conflict-Aware Instruction Fusion for Multi-Query Generative Engine Optimization

IF-GEO: Conflict-Aware Instruction Fusion for Multi-Query Generative Engine Optimization

Paper · 2026 · Heyang Zhou, Jiajia Chen, Xiaolu Chen et al.

Summary

Generative search engines answer queries by synthesizing retrieved sources, so content providers care about whether their document gets cited and visible in the generated answer. This is the goal of Generative Engine Optimization (GEO). Earlier GEO methods optimize toward one rule set or one aggregated intent. In practice, a single document must serve many heterogeneous queries, and their revision needs can clash within a limited content budget. IF-GEO addresses this with a "diverge-then-converge" pipeline built entirely from structured LLM calls:

  • Diverge phase: the LLM performs reverse retrieval to mine diverse, popularity-weighted latent queries. For each query, it then produces anchored edit requests with necessity scores.
  • Converge phase: conflict-aware instruction fusion prioritizes, deduplicates and reconciles these requests into a Global Revision Blueprint, which then guides the actual editing.

The authors also propose risk-aware stability metrics that capture tail degradations which mean and variance can hide. On GEO-Bench with GPT-4o-mini as the engine, IF-GEO beats strong baselines on overall gain and has lower downside risk. This shows that coordinating competing queries matters for robust GEO.

Key points

  • New stability metrics: besides Mean and VAR, the paper adds three measures: Downside Risk (DR), the Win-Tie Rate (WTR), and Worst-Case Performance (WCP).
  • Ablation: the full IF-GEO scores a Mean of 9.24 (WTR 80.80%, DR 0.0021). Scores fall without each component:
    • 8.18 without the blueprint;
    • 7.07 without instruction fusion;
    • 6.14 without conflict resolution, which also gives the worst WCP (−0.0713).
  • Number of expanded queries N: raising N from 1 to 9 lifts the Mean from 8.06 to 10.02. Stability gains level off after N=5, so N=5 is the default because cost grows roughly linearly with N.
  • Robustness: advantages hold when the engine is switched to Gemini-2.0-Flash and across documents with different initial ranking positions.
  • Limitations and code: the main limitations are higher token cost, reliance on LLM-simulated engines, and dependence on the quality of query discovery. Code is available at github.com/yangjizc/IF-GEO.

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 Science and Technology of ChinaHefei · China145 works · 114,409 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksInformation Systems · 1,730 works

Cite

Cite as Heyang Zhou, Jiajia Chen, Xiaolu Chen, Jie Bao, Zhen Chen and Yong Liao. 2026. IF-GEO: Conflict-Aware Instruction Fusion for Multi-Query Generative Engine Optimization. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1373

@inproceedings{zhou2026if,
  title = {{IF-GEO: Conflict-Aware Instruction Fusion for Multi-Query Generative Engine Optimization}},
  author = {Heyang Zhou, Jiajia Chen, Xiaolu Chen, Jie Bao, Zhen Chen and Yong Liao},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1373},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1373}
}

References

28 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, and Ameet Deshpande. 2024. GEO: generative engine optimization. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2024, Barcelona, Spain, August 25-29, 2024, pages 5–1… find
  2. Puneet S Bagga, Vivek F Farias, Tamar Korkotashvili, Tianyi Peng, and Yuhang Wu. 2025. E-geo: A testbed for generative engine optimization in e-commerce. arXiv preprint arXiv:2511.20867. — [arXiv:2511.20867](https://arxiv.org/abs/2511.20867) find
  3. Aharon Ben-Tal and Arkadi Nemirovski. 1998. Robust convex optimization. Mathematics of operations research, 23(4):769–805. find
  4. Sebastian Borgeaud, Arthur Mensch, Jordan Hoffmann, Trevor Cai, Eliza Rutherford, Katie Millican, George van den Driessche, Jean-Baptiste Lespiau, Bogdan Damoc, Aidan Clark, Diego de Las Casas, Aurelia Guy, Jacob Menick, Roman Ring, Tom Hennigan, Saffron Huang, Loren Maggiore, Chris Jones, Albin Cas… find
  5. Sergey Brin and Lawrence Page. 1998. The anatomy of a large-scale hypertextual web search engine. Comput. Networks, 30(1-7):107–117.
  6. Andrei Broder. 2002. A taxonomy of web search. SI-GIR Forum, 36(2):3–10. find
  7. Mahe Chen, Xiaoxuan Wang, Kaiwen Chen, and Nick Koudas. 2025a. Generative engine optimization: How to dominate AI search. CoRR, abs/2509.08919. — [arXiv:2509.08919](https://arxiv.org/abs/2509.08919) find
  8. Xiaolu Chen and Yong Liao. 2025. Caption injection for optimization in generative search engine. CoRR, abs/2511.04080. — [arXiv:2511.04080](https://arxiv.org/abs/2511.04080) find
  9. Xiaolu Chen, Haojie Wu, Jie Bao, Zhen Chen, Yong Liao, and Hu Huang. 2025b. Role-augmented intent-driven generative search engine optimization. CoRR, abs/2508.11158. — [arXiv:2508.11158](https://arxiv.org/abs/2508.11158) find
  10. Charles L. A. Clarke, Maheedhar Kolla, Gordon V. Cormack, Olga Vechtomova, Azin Ashkan, Stefan Büttcher, and Ian MacKinnon. 2008. Novelty and diversity in information retrieval evaluation. In Proceedings of the 31st Annual International ACM SI-GIR Conference on Research and Development in Informatio… find
  11. Tianyu Gao, Howard Yen, Jiatong Yu, and Danqi Chen. 2023. Enabling large language models to generate text with citations. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023, pages 6465–6488. Association for Computatio…
  12. Jeffrey L. Gleason, Desheng Hu, Ronald E. Robertson, and Christo Wilson. 2023. Google the gatekeeper: How search components affect clicks and attention. In Proceedings of the Seventeenth International AAAI Conference on Web and Social Media, ICWSM 2023, Limassol, Cyprus, June 5-8, 2023, pages 245–25… find
  13. Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Ming-Wei Chang. 2020. REALM: retrieval-augmented language model pre-training. volume abs/2002.08909.
  14. Gautier Izacard and Edouard Grave. 2021. Leveraging passage retrieval with generative models for open domain question answering. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, EACL 2021, Online, April 19 - 23, 2021, pages…
  15. Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. Dense passage retrieval for open-domain question answering. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, EMNLP 2020, Online, November…
  16. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. 2020. find
  17. Retrieval-augmented generation for knowledge-intensive NLP tasks. In Advances in Neural Information Processing Systems 33: Annual Conference on Neural Information Processing Systems 2020, NeurIPS 2020, December 6-12, 2020, virtual.
  18. Xiaoxi Li, Jiajie Jin, Yujia Zhou, Yuyao Zhang, Peitian Zhang, Yutao Zhu, and Zhicheng Dou. 2025. From matching to generation: A survey on generative information retrieval. ACM Trans. Inf. Syst., 43(3):83:1– 83:62. find
  19. Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. G-eval: NLG evaluation using gpt-4 with better human alignment. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023, pages 2511–2522.…
  20. Florian Lüttgenau, Imar Colic, and Gervasio Ramirez. 2025. Beyond SEO: A transformer-based approach for reinventing web content optimisation. CoRR, abs/2507.03169. — [arXiv:2507.03169](https://arxiv.org/abs/2507.03169) find
  21. Chokri Mamoghli and Sami Daboussi. 2009. Performance measurement of hedge funds portfolios in a downside risk framework. The Journal of Wealth Management, 12(2):101. find
  22. R Timothy Marler and Jasbir S Arora. 2004. Survey of multi-objective optimization methods for engineering. Structural and multidisciplinary optimization, 26(6):369–395. find
  23. Jacob Menick, Maja Trebacz, Vladimir Mikulik, John Aslanides, H. Francis Song, Martin J. Chadwick, Mia Glaese, Susannah Young, Lucy Campbell-Gillingham, Geoffrey Irving, and Nat McAleese. 2022. Teaching language models to support answers with verified quotes. CoRR, abs/2203.11147.
  24. Reiichiro Nakano, Jacob Hilton, Suchir Balaji, Jeff Wu, Long Ouyang, Christina Kim, Christopher Hesse, Shantanu Jain, Vineet Kosaraju, William Saunders, Xu Jiang, Karl Cobbe, Tyna Eloundou, Gretchen Krueger, Kevin Button, Matthew Knight, Benjamin Chess, and John Schulman. 2021. Webgpt: Browser-assis…
  25. Filip Radlinski, Madhu Kurup, and Thorsten Joachims. 2008. How does clickthrough data reflect retrieval quality? In Proceedings of the 17th ACM Conference on Information and Knowledge Management, CIKM 2008, Napa Valley, California, USA, October 26-30, 2008, pages 43–52. ACM. find
  26. Daniel E. Rose and Danny Levinson. 2004. Understanding user goals in web search. In Proceedings of the 13th international conference on World Wide Web, WWW 2004, New York, NY, USA, May 17-20, 2004, pages 13–19. ACM. find
  27. Xuanhui Wang, Deepayan Chakrabarti, and Kunal Punera. 2009. Mining broad latent query aspects from search sessions. In Proceedings of the 15th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Paris, France, June 28 - July 1, 2009, pages 867–876. ACM. find
  28. Yujiang Wu, Shanshan Zhong, Yubin Kim, and Chenyan Xiong. 2025. What generative search engines like and how to optimize web content cooperatively. CoRR, abs/2510.11438. — [arXiv:2510.11438](https://arxiv.org/abs/2510.11438) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhou-2026-if-geo-conflict-aware-instruction/