Large Language Models (LLMs) have made significant progress in open-ended dialogue, yet their inability to retain and retrieve relevant information from long-term interactions limits their effectiveness in applications requiring sustained personalization. External memory mechanisms have been proposed to address this limitation, enabling LLMs to maintain conversational continuity. However, existing approaches struggle with two key challenges. First, rigid memory granularity fails to capture the natural semantic structure of conversations, leading to fragmented and incomplete representations. Second, fixed retrieval mechanisms cannot adapt to diverse dialogue contexts and user interaction patterns. In this work, we propose Reflective Memory Management (RMM), a novel mechanism for long-term dialogue agents, integrating forward- and backward-looking reflections: (1) Prospective Reflection, which dynamically summarizes interactions across granularities—utterances, turns, and sessions—into a personalized memory bank for effective future retrieval, and (2) Retrospective Reflection, which iteratively refines the retrieval in an online reinforcement learning (RL) manner based on LLMs’ cited evidence. Experiments show that RMM demonstrates consistent improvement across various metrics and benchmarks. For example, RMM shows more than 10% accuracy improvement over the baseline without memory management on the LongMemEval dataset.

Fan Yin, Zifeng Wang, I-Hung Hsu, Jun Yan, Ke Jiang, Yanfei Chen, Jindong Gu, Long Le, Kai-Wei Chang, Chen-Yu Lee, Hamid Palangi, Tomas Pfister. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Mihir Parmar, Xin Liu, Palash Goyal, Yanfei Chen, Long Le, Swaroop Mishra, Hossein Mobahi, Jindong Gu, Zifeng Wang, Hootan Nakhost, Chitta Baral, Chen-Yu Lee, Tomas Pfister, Hamid Palangi. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Embeddings extracted by pre-trained Large Language Models (LLMs) have significant potential to improve information retrieval and search.Beyond the zero-shot setup in which they are being conventionally used, being able to take advantage of the information from the relevant query-corpus paired data can further boost the LLM capabilities.In this paper, we propose a novel method, Search-Adaptor, for customizing LLMs for information retrieval in an efficient and robust way.Search-Adaptor modifies the embeddings generated by pre-trained LLMs, and can be integrated with any LLM, including those only available via prediction APIs.On multiple English, multilingual, and multimodal retrieval datasets, we show consistent and significant performance benefits for Search-Adaptor -e.g., more than 5% improvements for Google Embedding APIs in nDCG@10 averaged over 14 BEIR datasets.

Recent advancements in language models (LMs) have notably enhanced their ability to reason with tabular data, primarily through program-aided mechanisms that manipulate and analyze tables. However, these methods often require the entire table as input, leading to scalability challenges due to the positional bias or context length constraints. In response to these challenges, we introduce TableRAG, a Retrieval-Augmented Generation (RAG) framework specifically designed for LM-based table understanding. TableRAG leverages query expansion combined with schema and cell retrieval to pinpoint crucial information before providing it to the LMs. This enables more efficient data encoding and precise retrieval, significantly reducing prompt lengths and mitigating information loss. We have developed two new million-token benchmarks from the Arcade and BIRD-SQL datasets to thoroughly evaluate TableRAG's effectiveness at scale. Our results demonstrate that TableRAG's retrieval design achieves the highest retrieval quality, leading to the new state-of-the-art performance on large-scale table understanding.

Addressing the challenge of effectively processing long contexts has become a critical issue for Large Language Models (LLMs). Two common strategies have emerged: 1) reducing the input length, such as retrieving relevant chunks by Retrieval-Augmented Generation (RAG), and 2) expanding the context window limit of LLMs. However, both strategies have drawbacks: input reduction has no guarantee of covering the part with needed information, while window extension struggles with focusing on the pertinent information for solving the task. To mitigate these limitations, we propose Chain-of-Agents (CoA), a novel framework that harnesses multi-agent collaboration through natural language to enable information aggregation and context reasoning across various LLMs over long-context tasks. CoA consists of multiple worker agents who sequentially communicate to handle different segmented portions of the text, followed by a manager agent who synthesizes these contributions into a coherent final output. CoA processes the entire input by interleaving reading and reasoning, and it mitigates long context focus issues by assigning each agent a short context. We perform a comprehensive evaluation of CoA on a wide range of long-context tasks in question answering, summarization, and code completion, demonstrating significant improvements by up to 10% over strong baselines of RAG, Full-Context, and multi-agent LLMs.

Voxel is one of the common structural representation of 3D point cloud. Due to the sparsity of point cloud generated by light detection and ranging (LiDAR), there is the extreme imbalance in the foreground and background voxels. It decreases the accuracy of 3D object detection, has the negative effect on intelligent driving safety. To overcome this problem, we present a saliency prediction based 3D object detector SP-Det in this article. Although foreground voxels have the sufficient feature of object, it is difficult to localize the foreground region from voxel space with the larger background region. We design an auxiliary learning task, saliency prediction (SP). It benefits 3D detector in identifying the foreground region. SP task uses label diffusion to alleviate the label imbalance. It reduces the learning difficulty of saliency in voxel and bird's eye view (BEV) spaces. After that, to strengthen feature interaction from the sparse foreground region, we design saliency fusion (SF) module to fuse the learning result in SP task. It utilizes voxel and BEV saliency maps as progressive attention to resist the redundant feature from background region. To aggregate more foreground feature inside 3D and BEV region of interest (RoI), we design hybrid grid maps based RoI pooling (Hybrid-RoI pooling). Experiments are conducted in STF dataset. The adverse weather enlarges the sparsity of LiDAR point cloud, increasing the difficulty of object detection. SP-Det identifies and leverages foreground region, and achieves the performance better than the current methods. Hence, we believe that SP-Det benefits to LiDAR based 3D scene understanding in the adverse weather.

Light detection and ranging (LiDAR) is an essential sensor for three dimensional (3D) object detection via generating 3D point cloud of the surroundings, and it has been widely used in the various visual applications, especially autonomous driving. However, limited numbers of labeled LiDAR datasets brutally restrain the development of 3D object detector, and this situation breeds an urgent demand on data augmentation in this field. By far, most of the traditional methods reuse the labeled samples, while those unlabeled are hastily untaken. Motivated by this, we propose aRealisticSimulator based data augmentation (RS-Aug). It aims to construct augmented real scenes to enrich the diversity of training dataset. To train 3D object detector in a supervised learning way, the first step of RS-Aug is auto-annotation. Time-continuous LiDAR frames are used to construct the dense scene, which is beneficial to annotation and the subsequent rendering augmentation. However, 3D points with incorrect semantic labels are naturally gathered during multi-view reconstruction, causing the negative effect on auto-annotation. We propose an algorithm of cluster guided$k$-nearest neighbor (c-$k$NN). It emphasizes on de-nosing semantic labels of clustered points using distance and intensity constraints. Then, the next step of RS-Aug is rendering augmentation on the real scene. To enhance the rendering quality using collision and distance constraints with the less computation complexity, we propose a scheme of heuristic search (HS) based object insertion. It estimates the proper position of the inserted object from 2D bird’s eye view (BEV). Experiments demonstrate the de-noising accuracy of c-$k$NN, rendering quality of HS based object insertion, and improvement of RS-Aug on object detection.

A locally adaptive shrinkage Bayesian estimate for medical ultrasonography denoising is proposed by exploiting the correlation among image sparse coding. The Laplacian distribution is used to model the coding coefficients. The paper deduces the MAP estimate formula and adaptive threshold. Simulation experiments are carried out to show the effectiveness of the new method. Results demonstrate that compared with classical denoising algorithms, the new method has increased peak signal-to-noise ratio (PSNR) and improved the quality of subjective visual effect. Our algorithm is also proved to be effective to the medical ultrasonography.

Ship target recognition in infrared image remains a difficult problem, due to the projection or silhouette of a three-dimensional ship target being variable in shape, orientation and scale to make its recognizability unstable. In this paper, a transductive transfer learning framework is proposed to solve the problem. Hu moments is firstly extracted as feature vectors of target. Then the transductive transfer learning method is used to find the common parameters between the feature spaces of the training ship samples and the detected ship targets, and transfer the similar knowledge from those data with different distributions. According to the experiment result in simulation infrared images, it shows that the ship targets can be recognized highly and reliably by our proposed framework. It demonstrates the robustness and effectiveness of our method for infrared images.