Efficient multi-hop reasoning requires Large Language Models (LLMs) based agents to acquire high-value external knowledge iteratively. Previous work has explored reinforcement learning (RL) to train LLMs to perform search-based document retrieval, achieving notable improvements in QA performance, but underperform on complex, multi-hop QA resulting from the sparse rewards from global signal only. To address this gap in existing research, we introduce StepSearch, a framework for search LLMs that trained with step-wise proximal policy optimization method. It consists of richer and more detailed intermediate search rewards and token-level process supervision based on information gain and redundancy penalties to better guide each search step. We constructed a fine-grained question-answering dataset containing sub-question-level search trajectories based on open source datasets through a set of data pipeline method. On standard multi-hop QA benchmarks, it significantly outperforms global-reward baselines, achieving 11.2% and 4.2% absolute improvements for 3B and 7B models over various search with RL baselines using only 19k training data, demonstrating the effectiveness of fine-grained, stepwise supervision in optimizing deep search LLMs. The project is open source at https://github.com/Zillwang/StepSearch

“Multimodal Named Entity Recognition (MNER) is a challenging task in social mediadue to the combination of text and image features. Previous MNER work has focused onpredicting entity information after fusing visual and text features. However, pre-traininglanguage models have already acquired vast amounts of knowledge during their pre-training process. To leverage this knowledge, we propose a prompt network for MNERtasks (P-MNER).To minimize the noise generated by irrelevant areas in the image, wedesign a visual feature extraction model (FRR) based on FasterRCNN and ResNet, whichuses fine-grained visual features to assist MNER tasks. Moreover, we introduce a textcorrection fusion module (TCFM) into the model to address visual bias during modalfusion. We employ the idea of a residual network to modify the fused features using theoriginal text features. Our experiments on two benchmark datasets demonstrate that ourproposed model outperforms existing MNER methods. P-MNER’s ability to leveragepre-training knowledge from language models, incorporate fine-grained visual features,and correct for visual bias, makes it a promising approach for multimodal named entityrecognition in social media posts.”