AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

HiPRAG: Hierarchical Process Rewards for Efficient Agentic Retrieval Augmented Generation

Peilin Wu, Mian Zhang, Kun Wan, Wentian Zhao, Kaiyu He, Xinya Du, Zhiyu Chen

ICLR 2026 · Large language models › Retrieval and knowledge · approach: RAG robustness and evaluation · 16 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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 changedWhereMeasuredEffect
add HiPRAG hierarchical process rewardQwen2.5-7B-Instruct GRPO · 7 QA benchmarksAvg. CEM: 61.2 -> 67.2Improved
add HiPRAG process rewardQwen2.5-7B-Instruct PPO · 7 QA benchmarks avgAvg. CEM: 53.3 -> 64.5Improved
add HiPRAG process rewardLlama-3.2-3B-Instruct PPO · 7 QA benchmarks avgAvg. CEM: 56.4 -> 64.8Improved
add HiPRAG process rewardQwen2.5-3B-Instruct GRPO · 7 QA benchmarks avgAvg. OSR: 8.4 -> 4.1Improved
use HiPRAG parsable step output format with Search-R1 outcome reward (ablation)Search-R1 · 7 QA benchmarks avgAvg. CEM: 60.3 -> 61.2No clear effect
use HiPRAG parsable step output format with beta-GRPO reward (ablation)beta-GRPO · 7 QA benchmarks avgAvg. CEM: 62.5 -> 62.1No clear effect
add over-search-only penalty instead of full process reward (ablation)HiPRAG Qwen2.5-3B-Instruct PPO · 7 QA benchmarks avgAvg. CEM: 64.1 -> 58.8Hurt
add over-search-only penalty instead of full process reward (ablation)HiPRAG Qwen2.5-3B-Instruct PPO · 7 QA benchmarks avgAvg. USR: 38.1 -> 52.7Hurt
add under-search-only penalty instead of full process reward (ablation)HiPRAG Qwen2.5-3B-Instruct PPO · 7 QA benchmarks avgAvg. CEM: 64.1 -> 63.3Hurt
add under-search-only penalty instead of full process reward (ablation)HiPRAG Qwen2.5-3B-Instruct PPO · 7 QA benchmarks avgAvg. OSR: 4.9 -> 6.6Hurt
process bonus coefficient 0.2 instead of 0.4 (ablation)HiPRAG Qwen2.5-3B-Instruct PPO · 7 QA benchmarks avgAvg. CEM: 64.1 -> 59.6Hurt
process bonus coefficient 0.6 instead of 0.4 (ablation)HiPRAG Qwen2.5-3B-Instruct PPO · 7 QA benchmarks avgAvg. CEM: 64.1 -> 62.5Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy