Natural Language Query to Configuration for Retrieval Agents
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913164653756416 |
|---|---|
| author | Pan, Melissa Z. Arabzadeh, Negar Jacob, Mathew Kazhamiaka, Fiodar Choukse, Esha Zaharia, Matei |
| author_facet | Pan, Melissa Z. Arabzadeh, Negar Jacob, Mathew Kazhamiaka, Fiodar Choukse, Esha Zaharia, Matei |
| contents | Modern retrieval agents expose many configuration choices -- LLM, retriever, number of documents, number of hops, and synthesis strategy -- each shaping both answer quality and serving cost. Today, these pipelines are typically hand-tuned once per workload, leaving substantial per-query optimization untapped. We formulate the problem: given a natural-language query and either an accuracy or a budget target, select from a predefined pipeline catalog the configuration that minimizes cost or maximizes accuracy at inference time. We propose **BRANE**, which uses an LLM to convert each query into workload-specific characteristics, then trains a lightweight per-configuration predictor that estimates whether the pipeline will answer the query correctly. At inference time, **BRANE** selects the configuration that maximizes predicted correctness penalized by cost, exposing a tunable cost-quality tradeoff without retraining. Across MuSiQue, BrowseComp-Plus, and FinanceBench, **BRANE** consistently pushes the cost-quality Pareto frontier, matches the best fixed configuration's accuracy at up to 89% lower cost, and outperforms LLM-routing, rule-based, and fine-tuned Qwen3-4B baselines. These results show that per-query configuration of the full retrieval pipeline is a practical alternative to static workload-level tuning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_27361 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Natural Language Query to Configuration for Retrieval Agents Pan, Melissa Z. Arabzadeh, Negar Jacob, Mathew Kazhamiaka, Fiodar Choukse, Esha Zaharia, Matei Artificial Intelligence Systems and Control Modern retrieval agents expose many configuration choices -- LLM, retriever, number of documents, number of hops, and synthesis strategy -- each shaping both answer quality and serving cost. Today, these pipelines are typically hand-tuned once per workload, leaving substantial per-query optimization untapped. We formulate the problem: given a natural-language query and either an accuracy or a budget target, select from a predefined pipeline catalog the configuration that minimizes cost or maximizes accuracy at inference time. We propose **BRANE**, which uses an LLM to convert each query into workload-specific characteristics, then trains a lightweight per-configuration predictor that estimates whether the pipeline will answer the query correctly. At inference time, **BRANE** selects the configuration that maximizes predicted correctness penalized by cost, exposing a tunable cost-quality tradeoff without retraining. Across MuSiQue, BrowseComp-Plus, and FinanceBench, **BRANE** consistently pushes the cost-quality Pareto frontier, matches the best fixed configuration's accuracy at up to 89% lower cost, and outperforms LLM-routing, rule-based, and fine-tuned Qwen3-4B baselines. These results show that per-query configuration of the full retrieval pipeline is a practical alternative to static workload-level tuning. |
| title | Natural Language Query to Configuration for Retrieval Agents |
| topic | Artificial Intelligence Systems and Control |
| url | https://arxiv.org/abs/2605.27361 |