Research map / Large language models / Retrieval and knowledge
OPERA: A Reinforcement Learning--Enhanced Orchestrated Planner-Executor Architecture for Reasoning-Oriented Multi-Hop Retrieval
Yu Liu, Yanbing Liu, Fangfang Yuan, Cong Cao, Youbang Sun, Kun Peng, WeiZhuo Chen, Jianjun Li, Zhiyuan Ma
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the Plan Agent (ablation) | OPERA · MuSiQue | EM: 17.1 -> 39.7 | Improved |
| add the Rewrite Agent (ablation) | OPERA · MuSiQue | EM: 34.5 -> 39.7 | Improved |
| keep both Plan and Rewrite Agents (ablation) | OPERA · MuSiQue | EM: 16.7 -> 39.7 | Improved |
| train with SFT instead of CoT prompting (ablation) | OPERA architecture · MuSiQue | EM: 21.2 -> 24.3 | Improved |
| train with GRPO instead of SFT (ablation) | OPERA architecture · MuSiQue | EM: 24.3 -> 34.8 | Improved |
| use MAPGRPO (role-specific rewards, sequential training) instead of standard GRPO (ablation) | OPERA architecture · MuSiQue | EM: 34.8 -> 39.7 | Improved |
| train with SFT instead of CoT (ablation) | OPERA architecture · NQ (single-hop, out-of-domain) | EM: 23.1 -> 19.5 | Hurt |
| train with MAPGRPO instead of CoT (ablation) | OPERA architecture · NQ (out-of-domain) | EM: 23.1 -> 36.6 | Improved |
| train with MAPGRPO instead of CoT (ablation) | OPERA architecture · MultiHopRAG (out-of-domain) | EM: 42.3 -> 55.7 | Improved |
| inject high-scoring expert samples into each GRPO group (25%) (ablation) | MAPGRPO training (7B) · gradient variance | policy gradient variance | Improved |
Closest papers by meaning
- Scent of Knowledge: Optimizing Search-Enhanced Reasoning with Information Foraging (NeurIPS 2025 · similarity 0.88)
- Retro*: Optimizing LLMs for Reasoning-Intensive Document Retrieval (ICLR 2026 · similarity 0.88)
- MC-Search: Evaluating and Enhancing Multimodal Agentic Search with Structured Long Reasoning Chains (ICLR 2026 · similarity 0.88)
- AceSearcher: Bootstrapping Reasoning and Search for LLMs via Reinforced Self-Play (NeurIPS 2025 · similarity 0.88)
- MMhops-R1: Multimodal Multi-hop Reasoning (AAAI 2026 · similarity 0.87)
- Search and Refine During Think: Facilitating Knowledge Refinement for Improved Retrieval-Augmented Reasoning (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
