DRIFT-BENCH: Diagnosing CoopeRative Breakdowns in LLM Agents under Input Faults via Multi-Turn Interaction

Han Bao, Zheyuan Zhang, Pengcheng Jing, Zhengqing Yuan, Kaiwen Shi, Yanfang Ye
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:6492-6559, 2026.

Abstract

As Large Language Models transition to autonomous agents, user inputs frequently violate cooperative assumptions (e.g., implicit intent, missing parameters, false presuppositions, or ambiguous expressions), creating execution risks that text-only evaluations do not capture. Existing benchmarks typically assume well-specified instructions or restrict evaluation to text-only, single-turn clarification, and thus do not measure multi-turn disambiguation under grounded execution risk. We introduce DRIFT-BENCH, the first diagnostic benchmark that evaluates agentic pragmatics under input faults through multi-turn clarification across state-oriented and service-oriented execution environments. Grounded in classical theories of communication, DRIFT-BENCH provides a unified taxonomy of cooperative breakdowns and employs a persona-driven user simulator with the Rise evaluation protocol. Experiments show substantial performance drops under these faults, with clarification effectiveness varying across user personas and fault types. DRIFT-BENCH connects clarification studies with agent benchmarking, providing a framework to diagnose failures arising from faulty user inputs.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-bao26e, title = {{DRIFT}-{BENCH}: Diagnosing {C}oope{R}ative Breakdowns in {LLM} Agents under Input Faults via Multi-Turn Interaction}, author = {Bao, Han and Zhang, Zheyuan and Jing, Pengcheng and Yuan, Zhengqing and Shi, Kaiwen and Ye, Yanfang}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {6492--6559}, 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/bao26e/bao26e.pdf}, url = {https://proceedings.mlr.press/v306/bao26e.html}, abstract = {As Large Language Models transition to autonomous agents, user inputs frequently violate cooperative assumptions (e.g., implicit intent, missing parameters, false presuppositions, or ambiguous expressions), creating execution risks that text-only evaluations do not capture. Existing benchmarks typically assume well-specified instructions or restrict evaluation to text-only, single-turn clarification, and thus do not measure multi-turn disambiguation under grounded execution risk. We introduce DRIFT-BENCH, the first diagnostic benchmark that evaluates agentic pragmatics under input faults through multi-turn clarification across state-oriented and service-oriented execution environments. Grounded in classical theories of communication, DRIFT-BENCH provides a unified taxonomy of cooperative breakdowns and employs a persona-driven user simulator with the Rise evaluation protocol. Experiments show substantial performance drops under these faults, with clarification effectiveness varying across user personas and fault types. DRIFT-BENCH connects clarification studies with agent benchmarking, providing a framework to diagnose failures arising from faulty user inputs.} }
Endnote
%0 Conference Paper %T DRIFT-BENCH: Diagnosing CoopeRative Breakdowns in LLM Agents under Input Faults via Multi-Turn Interaction %A Han Bao %A Zheyuan Zhang %A Pengcheng Jing %A Zhengqing Yuan %A Kaiwen Shi %A Yanfang Ye %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-bao26e %I PMLR %P 6492--6559 %U https://proceedings.mlr.press/v306/bao26e.html %V 306 %X As Large Language Models transition to autonomous agents, user inputs frequently violate cooperative assumptions (e.g., implicit intent, missing parameters, false presuppositions, or ambiguous expressions), creating execution risks that text-only evaluations do not capture. Existing benchmarks typically assume well-specified instructions or restrict evaluation to text-only, single-turn clarification, and thus do not measure multi-turn disambiguation under grounded execution risk. We introduce DRIFT-BENCH, the first diagnostic benchmark that evaluates agentic pragmatics under input faults through multi-turn clarification across state-oriented and service-oriented execution environments. Grounded in classical theories of communication, DRIFT-BENCH provides a unified taxonomy of cooperative breakdowns and employs a persona-driven user simulator with the Rise evaluation protocol. Experiments show substantial performance drops under these faults, with clarification effectiveness varying across user personas and fault types. DRIFT-BENCH connects clarification studies with agent benchmarking, providing a framework to diagnose failures arising from faulty user inputs.
APA
Bao, H., Zhang, Z., Jing, P., Yuan, Z., Shi, K. & Ye, Y.. (2026). DRIFT-BENCH: Diagnosing CoopeRative Breakdowns in LLM Agents under Input Faults via Multi-Turn Interaction. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:6492-6559 Available from https://proceedings.mlr.press/v306/bao26e.html.

Related Material