---
title: "Evidence-Augmented Generation Reasoning for Extremely"
description: "The authors therefore propose a two-stage, training-free pipeline: Dynamic Knowledge Construction: an LLM (DeepSeek-R1)… 18 references."
authors: ["Xiaoyu Zhu", "Long Yuan", "Rui Qi", "Jinan Xu"]
year: 2026
venue: "Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026)"
source: "https://aclanthology.org/2026.mellm-1.2/"
doi: "10.18653/v1/2026.mellm-1.2"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/zhu-2026-evidence-augmented-generation-reasoning-extremely/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/zhu-2026-evidence-augmented-generation-reasoning-extremely/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/32/) / Evidence-Augmented Generation Reasoning for Extremely Low-Resource Language Decipherment

# Evidence-Augmented Generation Reasoning for Extremely Low-Resource Language Decipherment

**Authors:** Xiaoyu Zhu, [Long Yuan](https://md.research.elamur.ai/free/en/authors/long-yuan/), [Rui Qi](https://md.research.elamur.ai/free/en/authors/rui-qi/) and [Jinan Xu](https://md.research.elamur.ai/free/en/authors/jinan-xu/) · **Year:** 2026 · **Venue:** [Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026)](https://md.research.elamur.ai/free/en/venues/mellm/) ([Workshop on Multilinguality in the Era of Large Language Models 2026](https://md.research.elamur.ai/free/en/venues/mellm/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.mellm-1.2/) · [PDF](https://aclanthology.org/2026.mellm-1.2.pdf) · [DOI 10.18653/v1/2026.mellm-1.2](https://doi.org/10.18653/v1/2026.mellm-1.2) · [OpenAlex](https://openalex.org/W7166803607)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper addresses Rosetta Stone-style puzzles from linguistic Olympiads. A model sees a few sentence pairs in an extremely low-resource or extinct language plus English and must infer grammar and vocabulary to translate new items, match words, or fill in forms. Plain few-shot or chain-of-thought prompting tends to capture only surface regularities. The authors therefore propose a two-stage, training-free pipeline:

1. **Dynamic Knowledge Construction:** an LLM (DeepSeek-R1) produces many synthetic examples, each embodying a hypothesised rule of the target language.
2. **Task-Aware Evidence Augmentation:** a retriever selects the top-K most relevant examples as evidence for each query. This keeps the context small and supports analogical reasoning.

The design is motivated by cognitive load theory and analogical reasoning theory. On the LINGOLY and Linguini benchmarks, the method is reported to beat existing baselines across open-source and commercial LLMs of different sizes. The work matters because it reframes decipherment as on-the-fly rule induction rather than reliance on memorised knowledge. That framing is relevant to endangered-language work.

## Key points
- The evaluation draws on a pair of benchmarks constructed from low-resource or artificial languages, namely Linguini, introduced by Sánchez et al. (2024), and LINGOLY, introduced by Bean et al. (2024).
- Task types include sentence translation, numbered-rule problems (e.g., Roon numerals), and word matching (e.g., Hmong).
- Prior methods compared against include few-shot/CoT prompting and analogical prompting (Ramji and Ramji, 2025). The latter generates exemplars in related higher-resource languages and was applied with GPT-4o and Llama-3.1-405B.
- Synthetic exemplars serve only as in-context evidence, never as fine-tuning data. Quality control combines diversity filtering with a rule consistency check, and BM25 RAG appears as the retrieval variant in the case studies.
- In case studies on Basque and Yidiny, DeepSeek-R1 and DeepSeek-R1-Distill-Qwen-32B returned null answers with the original context alone but produced answers once retrieved exemplars were added. For Basque, these included the forms "ukitu nauzu" and "hurbildu zaizkit". A stated limitation is that biases of the generator propagate into the evidence pool.

*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

- [Beijing Jiaotong University](https://md.research.elamur.ai/free/en/universities/beijing-jiaotong-university/) — University, Beijing, China · 28 works · 1,754 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (2): [🧮 Reasoning](https://md.research.elamur.ai/free/en/topics/reasoning/) · 1,685 works · [🌱 Low-resource languages](https://md.research.elamur.ai/free/en/topics/low-resource-languages/) · 287 works

## Cite

Cite as: Xiaoyu Zhu, Long Yuan, Rui Qi and Jinan Xu. 2026. Evidence-Augmented Generation Reasoning for Extremely Low-Resource Language Decipherment. Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{zhu2026evidence,
  title = {{Evidence-Augmented Generation Reasoning for Extremely Low-Resource Language Decipherment}},
  author = {Xiaoyu Zhu, Long Yuan, Rui Qi and Jinan Xu},
  booktitle = {Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.mellm-1.2},
  url = {https://doi.org/10.18653/v1/2026.mellm-1.2}
}
```

- [← previous Evidence Grounding vs. Memorization: Why Neural Semantics Matter for Knowledge Graph Fact Verification](https://md.research.elamur.ai/free/en/papers/upadhyay-2026-evidence-grounding-vs-memorization-why/)
- [next → Evo-Attacker: Memory-Augmented Reinforcement Learning for Long-Horizon Tool Attacks on LLM-MAS](https://md.research.elamur.ai/free/en/papers/yan-2026-evo-attacker-memory-augmented-reinforcement/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/32/)

## References

18 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Rohan Anil, Andrew M. Dai, Orhan Firat, Melvin Johnson, Dmitry Lepikhin, Alexandre Passos, Siamak Shakeri, Emanuel Taropa, Paige Bailey, Zhifeng Chen, Eric Chu, Jonathan H. Clark, Laurent El Shafey, Yanping Huang, Kathy Meier-Hellstern, Gaurav Mishra, Erica Moreira, Mark Omernick, Kevin Robinson, an… — [PaLM 2 Technical Report](https://md.research.elamur.ai/free/en/papers/anil-2023-palm-2-technical-report/)
2. 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. [Retrieval-augmented generation for knowledge-intensive NLP tasks](https://md.research.elamur.ai/free/en/papers/lewis-2020-retrieval-augmented-generation-knowledge-intensive/). In Advances in Neural Information Proc…
3. Andrew M. Bean, Simi Hellsten, Harry Mayne, Jabez Magomere, Ethan Chi, Ryan Chi, Scott Hale, and Hannah Rose Kirk. 2024. LINGOLY: A benchmark of olympiad-level linguistic reasoning puzzles in low resource and extinct languages. In Advances in Neural Information Processing Systems 38: Annual Conferen… [find](https://scholar.google.com/scholar?q=Andrew+M%2E+Bean%2C+Simi+Hellsten%2C+Harry+Mayne%2C+Jabez+Magomere%2C+Ethan+Chi%2C+Ryan+Chi%2C+Scott+Hale%2C+and+Hannah+Rose+Kirk%2E+2024%2E+LINGOLY%3A+A+benchmark+of+olympiad%2Dlevel+linguistic+reasoning+puzzles+in+low+reso)
4. Yinheng Li. 2023. A practical survey on zero-shot prompt design for in-context learning. In Proceedings of the 14th International Conference on Recent Advances in Natural Language Processing, RANLP 2023, Varna, Bulgaria, 4-6 September 2023, pages 641–647. INCOMA Ltd., Shoumen, Bulgaria. [find](https://scholar.google.com/scholar?q=Yinheng+Li%2E+2023%2E+A+practical+survey+on+zero%2Dshot+prompt+design+for+in%2Dcontext+learning%2E+In+Proceedings+of+the+14th+International+Conference+on+Recent+Advances+in+Natural+Language+Processing%2C+RANLP+20)
5. Chin-Yew Lin. 2004. [Rouge: A package for automatic evaluation of summaries](https://md.research.elamur.ai/free/en/papers/lin-2004-rouge-package-automatic-evaluation-summaries/). In Text summarization branches out, pages 74–81.
6. Bozhidar Bozhanov and Ivan Derzhanski. 2013. Rosetta stone linguistic problems. In Proceedings of the Fourth Workshop on Teaching NLP and CL, pages 1–8, Sofia, Bulgaria. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Bozhidar+Bozhanov+and+Ivan+Derzhanski%2E+2013%2E+Rosetta+stone+linguistic+problems%2E+In+Proceedings+of+the+Fourth+Workshop+on+Teaching+NLP+and+CL%2C+pages+1%E2%80%938%2C+Sofia%2C+Bulgaria%2E+Association+for+Computational+)
7. Man Luo, Xin Xu, Yue Liu, Panupong Pasupat, and Mehran Kazemi. 2024. [In-context learning with retrieved demonstrations for language models: A survey](https://md.research.elamur.ai/free/en/papers/luo-2024-context-learning-retrieved-demonstrations-language/).
8. Trans. Mach. Learn. Res., 2024. [find](https://scholar.google.com/scholar?q=Trans%2E+Mach%2E+Learn%2E+Res%2E%2C+2024%2E)
9. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter,… [find](https://scholar.google.com/scholar?q=Tom+B%2E+Brown%2C+Benjamin+Mann%2C+Nick+Ryder%2C+Melanie+Subbiah%2C+Jared+Kaplan%2C+Prafulla+Dhariwal%2C+Arvind+Neelakantan%2C+Pranav+Shyam%2C+Girish+Sastry%2C+Amanda+Askell%2C+Sandhini+Agarwal%2C+Ariel+Herbert%2DVoss%2C+Gretche)
10. Language Raymond J. Mooney. 1996. Inductive logic programming for natural language processing. In Inductive Logic Programming, 6th International Workshop, ILP-96, Stockholm, Sweden, August 26-28, 1996, Selected Papers, volume 1314 of Lecture Notes in Computer Science, pages 3–22. Springer. [find](https://scholar.google.com/scholar?q=Language+Raymond+J%2E+Mooney%2E+1996%2E+Inductive+logic+programming+for+natural+language+processing%2E+In+Inductive+Logic+Programming%2C+6th+International+Workshop%2C+ILP%2D96%2C+Stockholm%2C+Sweden%2C+August+26%2D28%2C+1996)
11. Stephen H. Muggleton. 1991. Inductive logic programming. New Gener. Comput., 8(4):295–318. elicits reasoning in large language models. In Advances in Neural Information Processing Systems 35: Annual Conference on Neural Information Processing Systems 2022, NeurIPS 2022, New Orleans, LA, USA, Novembe… [find](https://scholar.google.com/scholar?q=Stephen+H%2E+Muggleton%2E+1991%2E+Inductive+logic+programming%2E+New+Gener%2E+Comput%2E%2C+8%284%29%3A295%E2%80%93318%2E+elicits+reasoning+in+large+language+models%2E+In+Advances+in+Neural+Information+Processing+Systems+35%3A+Annual+C)
12. Sam Musker, Alex Duchnowski, Raphaël Millière, and Ellie Pavlick. 2024. Llms as models for analogical reasoning. arXiv preprint arXiv:2406.13803. [find](https://scholar.google.com/scholar?q=Sam+Musker%2C+Alex+Duchnowski%2C+Rapha%C3%ABl+Milli%C3%A8re%2C+and+Ellie+Pavlick%2E+2024%2E+Llms+as+models+for+analogical+reasoning%2E+arXiv+preprint+arXiv%3A2406%2E13803%2E)
13. Xuanwang Zhang, Yunze Song, Yidong Wang, Shuyun Tang, Xinfeng Li, Zhengran Zeng, Zhen Wu, Wei Ye, Wenyuan Xu, Yue Zhang, Xinyu Dai, Shikun Zhang, and Qingsong Wen. 2024. RAGLAB: A modular and research-oriented unified framework for retrieval-augmented generation. CoRR, abs/2408.11381. [find](https://scholar.google.com/scholar?q=Xuanwang+Zhang%2C+Yunze+Song%2C+Yidong+Wang%2C+Shuyun+Tang%2C+Xinfeng+Li%2C+Zhengran+Zeng%2C+Zhen+Wu%2C+Wei+Ye%2C+Wenyuan+Xu%2C+Yue+Zhang%2C+Xinyu+Dai%2C+Shikun+Zhang%2C+and+Qingsong+Wen%2E+2024%2E+RAGLAB%3A+A+modular+and+research)
14. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. [Bleu: a method for automatic evaluation of machine translation](https://md.research.elamur.ai/free/en/papers/papineni-2002-bleu-method-automatic-evaluation-machine/). In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, July 6-12, 2002, Philadelphia, PA, USA, pages 311–318. ACL.
15. Penghao Zhao, Hailin Zhang, Qinhan Yu, Zhengren Wang, Yunteng Geng, Fangcheng Fu, Ling Yang, Wentao Zhang, and Bin Cui. 2024. [Retrieval-augmented generation for ai-generated content: A survey](https://md.research.elamur.ai/free/en/papers/zhao-2024-retrieval-augmented-generation-ai-generated/). CoRR, abs/2402.19473.
16. Maja Popovic. 2015. [chrf: character n-gram f-score for automatic MT evaluation](https://md.research.elamur.ai/free/en/papers/popovic-2015-chrf-character-n-gram-f/). In Proceedings of the Tenth Workshop on Statistical Machine Translation, WMT@EMNLP 2015, 17-18 September 2015, Lisbon, Portugal, pages 392–395. The Association for Computer Linguistics.
17. Hattie Zhou, Azade Nova, Hugo Larochelle, Aaron C. Courville, Behnam Neyshabur, and Hanie Sedghi. 2022. [Teaching algorithmic reasoning via in-context learning](https://md.research.elamur.ai/free/en/papers/zhou-2022-teaching-algorithmic-reasoning-context-learning/). CoRR, abs/2211.09066.
18. Luc De Raedt. 2008. Logical and relational learning. In Advances in Artificial Intelligence - SBIA 2008, 19th Brazilian Symposium on Artificial Intelligence, Savador, Brazil, October 26-30, 2008. Proceedings, volume 5249 of Lecture Notes in Computer Science, page 1. Springer. [find](https://scholar.google.com/scholar?q=Luc+De+Raedt%2E+2008%2E+Logical+and+relational+learning%2E+In+Advances+in+Artificial+Intelligence+%2D+SBIA+2008%2C+19th+Brazilian+Symposium+on+Artificial+Intelligence%2C+Savador%2C+Brazil%2C+October+26%2D30%2C+2008%2E+Proc)

*Updated 2 November 2026*