Hengxin Fun

Shengyao Zhuang, Yuansheng Ni, Hengxin Fun et al.

Retrieval for search agents is still inherited from non-agentic information retrieval: a retriever ranks the corpus and the agent reads a small set of returned documents. Recent direct corpus interaction (DCI) work shows that agents can instead interact with the raw corpus through shell tools such as grep and file reads. But unbounded interaction does not scale: every broad shell command is a scan over the whole corpus, and latency degrades sharply as the corpus grows. We argue that the role of retrieval for agentic search is not just to select documents that fit in the LLM context window, but to construct an interaction space: a bounded subset of the corpus the agent can explore with associated tools. Two design consequences follow. The space needs a boundary supplied by retrieval, and the objects within it should be processed for interaction. As a proof of concept, we propose RISE (Retrieving Interaction SpacE): we use BM25 to construct the interaction space; meanwhile, its documents are processed during indexing for shell-style navigation. On BrowseComp-Plus, RISE matches the pure-shell DCI baseline at 78% accuracy with gpt-5.4-mini at roughly one quarter of the per-query cost. At 1M documents, RISE-BM25 reaches 81% on gpt-5.4-mini, whereas DCI on gpt-5.4-nano degrades to 60% with 33 of 100 wall-clock failures.

Hengxin Fun, Sunil Gandhi, Sujith Ravi

Recently, there have been significant advances in neural methods for tackling knowledge-intensive tasks such as open domain question answering (QA). These advances are fueled by combining large pre-trained language models with learnable retrieval of documents. Majority of these models use separate encoders for learning query representation, passage representation for the retriever and an additional encoder for the downstream task. Using separate encoders for each stage/task occupies a lot of memory and makes it difficult to scale to a large number of tasks. In this paper, we propose a novel Retrieval Optimized Multi-task (ROM) framework for jointly training self-supervised tasks, knowledge retrieval, and extractive question answering. Our ROM approach presents a unified and generalizable framework that enables scaling efficiently to multiple tasks, varying levels of supervision, and optimization choices such as different learning schedules without changing the model architecture. It also provides the flexibility of changing the encoders without changing the architecture of the system. Using our framework, we achieve comparable or better performance than recent methods on QA, while drastically reducing the number of parameters.