An Interactive Paradigm for Deep Research

Lin Ai, Victor Bursztyn, Xiang Chen, Julia Hirschberg, Saayan Mitra
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:1242-1270, 2026.

Abstract

Recent advances in large language models (LLMs) have enabled deep research systems that synthesize comprehensive, report-style answers to open-ended queries by combining retrieval, reasoning, and generation. Yet, most frameworks rely on rigid workflows with one-shot scoping and long autonomous runs, offering little room for course correction if user intent shifts mid-process. We present SteER, a framework for steerable deep research that introduces interpretable, mid-process control into long-horizon research workflows. At each decision point, SteER uses a cost–benefit formulation to determine whether to pause for user input or proceed autonomously. It combines diversity-aware planning with utility signals that reward alignment, novelty, and coverage, and maintains a live persona model that evolves throughout the session. SteER outperforms state-of-the-art open-source and proprietary baselines by up to 22.80% on alignment, leads on quality metrics such as breadth and balance, and is preferred by human readers in 85%+ of pairwise alignment judgments. We also introduce a persona–query benchmark and data-generation pipeline. To our knowledge, this is the first work to advance deep research with an interactive, interpretable control paradigm, paving the way for controllable, user-aligned agents in long-form tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-ai26a, title = {An Interactive Paradigm for Deep Research}, author = {Ai, Lin and Bursztyn, Victor and Chen, Xiang and Hirschberg, Julia and Mitra, Saayan}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {1242--1270}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/ai26a/ai26a.pdf}, url = {https://proceedings.mlr.press/v306/ai26a.html}, abstract = {Recent advances in large language models (LLMs) have enabled deep research systems that synthesize comprehensive, report-style answers to open-ended queries by combining retrieval, reasoning, and generation. Yet, most frameworks rely on rigid workflows with one-shot scoping and long autonomous runs, offering little room for course correction if user intent shifts mid-process. We present SteER, a framework for steerable deep research that introduces interpretable, mid-process control into long-horizon research workflows. At each decision point, SteER uses a cost–benefit formulation to determine whether to pause for user input or proceed autonomously. It combines diversity-aware planning with utility signals that reward alignment, novelty, and coverage, and maintains a live persona model that evolves throughout the session. SteER outperforms state-of-the-art open-source and proprietary baselines by up to 22.80% on alignment, leads on quality metrics such as breadth and balance, and is preferred by human readers in 85%+ of pairwise alignment judgments. We also introduce a persona–query benchmark and data-generation pipeline. To our knowledge, this is the first work to advance deep research with an interactive, interpretable control paradigm, paving the way for controllable, user-aligned agents in long-form tasks.} }
Endnote
%0 Conference Paper %T An Interactive Paradigm for Deep Research %A Lin Ai %A Victor Bursztyn %A Xiang Chen %A Julia Hirschberg %A Saayan Mitra %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-ai26a %I PMLR %P 1242--1270 %U https://proceedings.mlr.press/v306/ai26a.html %V 306 %X Recent advances in large language models (LLMs) have enabled deep research systems that synthesize comprehensive, report-style answers to open-ended queries by combining retrieval, reasoning, and generation. Yet, most frameworks rely on rigid workflows with one-shot scoping and long autonomous runs, offering little room for course correction if user intent shifts mid-process. We present SteER, a framework for steerable deep research that introduces interpretable, mid-process control into long-horizon research workflows. At each decision point, SteER uses a cost–benefit formulation to determine whether to pause for user input or proceed autonomously. It combines diversity-aware planning with utility signals that reward alignment, novelty, and coverage, and maintains a live persona model that evolves throughout the session. SteER outperforms state-of-the-art open-source and proprietary baselines by up to 22.80% on alignment, leads on quality metrics such as breadth and balance, and is preferred by human readers in 85%+ of pairwise alignment judgments. We also introduce a persona–query benchmark and data-generation pipeline. To our knowledge, this is the first work to advance deep research with an interactive, interpretable control paradigm, paving the way for controllable, user-aligned agents in long-form tasks.
APA
Ai, L., Bursztyn, V., Chen, X., Hirschberg, J. & Mitra, S.. (2026). An Interactive Paradigm for Deep Research. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:1242-1270 Available from https://proceedings.mlr.press/v306/ai26a.html.

Related Material