ICR² Accepted to ACL 2025 Findings

Our paper on Eliciting In-context Retrieval and Reasoning for Long-context LLMs has been accepted into ACL 2025 Findings.

Apple Machine Learning Research.

Happy to announce our paper Eliciting In-context Retrieval and Reasoning for Long-context Large Language Models by Yifu Qiu, Varun Embar, Yizhe Zhang, Navdeep Jaitly, Shay Cohen, Benjamin Han is accepted into ACL 2025 Findings [1]! We have also updated the paper on arXiv to include additional results on LLaMA-3-Instruct-8B.

Unfortunately I won’t be in Vienna for the conference due to work, but if you are going, make sure to attend Yifu and Varun’s presentation!

This is a follow-up to the original ICR² preprint announcement from January. See also a follow-on observation about embedding-based retrieval limits where ICR² is referenced as a way out.

---

References

[1] Qiu, Yifu, Varun Embar, Yizhe Zhang, Navdeep Jaitly, Shay Cohen, and Benjamin Han. “Eliciting In-context Retrieval and Reasoning for Long-context Large Language Models.” arXiv, 2025. https://arxiv.org/abs/2501.08248 · Apple Machine Learning Research: https://machinelearning.apple.com/research/eliciting-in-context · Repo: https://github.com/apple/ml-icr2

Originally posted on LinkedIn.