MARS: Training Multilingual Search Agents for Retriever-Free Evidence Acquisition

Researchers posted arXiv paper 2611.09090v1 introducing MARS, a multilingual search agent that operates in direct corpus interaction (DCI) and uses reinforcement learning instead of learned retrievers. MARS adds corpus-native entity grounding to recover lexical anchors from corpus evidence and a redundancy-aware search reward to prevent training collapse in multilingual RL. The authors report MARS consistently outperforms strong retrieval-based and DCI baselines, with robust cross-lingual generalization and substantially fewer redundant corpus interactions.

arXiv:2611.09090v1 Announce Type: new Abstract: Existing multilingual retrieval-augmented generation relies heavily on learned retrievers and similarity-based ranking from multilingual collections, which are vulnerable to cross-lingual retrieval errors and indexing overhead. In this paper, we explore retriever-free multilingual search, where the model directly interacts with raw corpora through executable shell commands. This paradigm shifts the bottleneck to multilingual search action generation, where exact search is sensitive to entity surface forms and outcome-only supervision provides limited guidance for multi-turn search. To address these challenges, we propose MARS, a multilingual search agent in direct corpus interaction DCI with reinforcement learning RL . MARS introduces corpus-native entity grounding to recover reliable lexical anchors from corpus evidence, mitigating failure to retrieve evidence. Meanwhile, we present a redundancy-aware search reward to encourage concise and effective search trajectories, addressing training collapse in RL of multilingual scenarios. Experiments show that \ours~consistently outperforms strong retrieval-based and DCI baselines, while demonstrating robust cross-lingual generalization and substantially reducing redundant corpus interactions.