Jiazheng Kang

Jiazheng Kang, Bowen Zhang, Zixin Song et al.

ReAct-style agents for search-intensive, multi-step reasoning tasks rely largely on their own internal judgment to decide what evidence to seek, which reasoning or action step to take next, and when to stop, often producing shallow, redundant, or poorly targeted trajectories. Prior work has explored rubrics as external quality signals, but existing uses are mostly evaluative rather than action-guiding: rubrics typically serve as training-time rewards or post-hoc evaluators of completed outputs, and in deep-research settings they are often coarse-grained and report-level rather than step-level. We introduce Co-ReAct, a rubric-guided action-selection framework that uses rubrics as step-level guidance during inference. At each decision step, Co-ReAct injects a rubric into the agent's context to guide the next Reason-or-Act decision, specifying what the agent should target in evidence seeking, search, reasoning, or self-evaluation. To make this guidance reliable, we train a dedicated rubric generator with GRPO. Unlike prior pairwise or binary preference formulations, our objective optimizes a list-wise Spearman rank-correlation reward against multi-judge expert consensus rankings, encouraging rubrics that are discriminative rather than merely plausible. On DeepResearchBench and SQA-CS-V2, Co-ReAct consistently improves over ReAct and representative test-time compute baselines across search agents built on both 8B/14B open-source and frontier closed-source base models. The trained rubric generator can also serve as a drop-in component that improves these baselines without changing their underlying decision mechanisms. Our code is publicly available at https://github.com/ZBWpro/Co-ReAct.

Jiangwang Chen, Bowen Zhang, Zixin Song et al.

Although large language model (LLM) conversational systems process millions of multi-turn dialogues daily, they remain fundamentally reactive: they respond only after the user types a query. A key step toward proactive interaction is next-query prediction, which anticipates the user's subsequent query based solely on the preceding dialogue. Progress on this task is hindered by the lack of dedicated benchmarks and a fundamental efficiency--quality trade-off: naively concatenating full dialogue history incurs linearly growing token consumption, while truncating to the latest turn discards crucial cross-turn context. Our key insight is that accurate prediction does not require re-reading raw history; it suffices to track the user's evolving intent trajectory across topics, unresolved needs, and interest shifts. We propose OnePred, which maintains a recursively updated memory as its sole cross-turn context, bounding the per-turn cost independently of conversation length. We train the model via a two-stage reinforcement learning pipeline that first teaches what to predict, then what to compress, shaping the memory into a prediction-oriented intent chain. To establish a rigorous testbed, we introduce NQP-Bench, spanning three diverse subsets. Experiments demonstrate that OnePred reduces per-turn token consumption by up to 22$\times$ compared to full-history inputs while consistently exceeding all baselines in prediction quality, with larger gains on longer conversations. Our code is publicly available at https://github.com/ZBWpro/OnePred.

Jiazheng Kang, Mingming Ji, Zhe Zhao et al.

Large Language Models (LLMs) face a crucial challenge from fixed context windows and inadequate memory management, leading to a severe shortage of long-term memory capabilities and limited personalization in the interactive experience with AI agents. To overcome this challenge, we innovatively propose a Memory Operating System, i.e., MemoryOS, to achieve comprehensive and efficient memory management for AI agents. Inspired by the memory management principles in operating systems, MemoryOS designs a hierarchical storage architecture and consists of four key modules: Memory Storage, Updating, Retrieval, and Generation. Specifically, the architecture comprises three levels of storage units: short-term memory, mid-term memory, and long-term personal memory. Key operations within MemoryOS include dynamic updates between storage units: short-term to mid-term updates follow a dialogue-chain-based FIFO principle, while mid-term to long-term updates use a segmented page organization strategy. Our pioneering MemoryOS enables hierarchical memory integration and dynamic updating. Extensive experiments on the LoCoMo benchmark show an average improvement of 49.11% on F1 and 46.18% on BLEU-1 over the baselines on GPT-4o-mini, showing contextual coherence and personalized memory retention in long conversations. The implementation code is open-sourced at https://github.com/BAI-LAB/MemoryOS.

Ting Bai, Jiazheng Kang, Jiayang Fan

We introduce a comprehensive large-scale role-playing agent corpus, termed BaiJia, that comprises various Chinese historical characters. This corpus is noteworthy for being the pioneering compilation of low-resource data that can be utilized in large language models (LLMs) to engage in AI-driven historical role-playing agents. BaiJia addresses the challenges in terms of fragmented historical textual records in different forms and modalities, integrating various characters' information, including their biographical, literary, family relations, historical events, and so on. We conduct extensive experiments to demonstrate the effectiveness of our BaiJia agent corpus in bolstering the role-playing abilities of various foundational LLMs, and promoting the development and assessment of LLMs in the context of historical role-playing tasks. The agent corpus is available at baijia.online.

Jiazheng Kang, Le Huang, Cheng Hou et al.

In real-world industrial settings, large language models (LLMs) must learn continually to keep pace with diverse and evolving tasks, requiring self-evolution to refine knowledge under dynamic data distributions. However, existing continual learning (CL) approaches, such as replay and parameter isolation, often suffer from catastrophic forgetting: training on new tasks degrades performance on earlier ones by overfitting to the new distribution and weakening generalization.We propose MoE-CL, a parameter-efficient adversarial mixture-of-experts framework for industrial-scale, self-evolving continual instruction tuning of LLMs. MoE-CL uses a dual-expert design: (1) a dedicated LoRA expert per task to preserve task-specific knowledge via parameter independence, mitigating forgetting; and (2) a shared LoRA expert to enable cross-task transfer. To prevent transferring task-irrelevant noise through the shared pathway, we integrate a task-aware discriminator within a GAN. The discriminator encourages the shared expert to pass only task-aligned information during sequential training. Through adversarial learning, the shared expert acquires generalized representations that mimic the discriminator, while dedicated experts retain task-specific details, balancing knowledge retention and cross-task generalization and thereby supporting self-evolution.Extensive experiments on the public MTL5 benchmark and an industrial Tencent3 benchmark validate the effectiveness of MoE-CL for continual instruction tuning. In real-world A/B testing for content compliance review on the Tencent Video platform, MoE-CL reduced manual review costs by 15.3%. These results demonstrate that MoE-CL is practical for large-scale industrial deployment where continual adaptation and stable transfer are critical.

Xinlong Zhai, Chunchen Wang, Ruijia Wang et al.

Drug-target interaction prediction (DTI) is essential in various applications including drug discovery and clinical application. There are two perspectives of input data widely used in DTI prediction: Intrinsic data represents how drugs or targets are constructed, and extrinsic data represents how drugs or targets are related to other biological entities. However, any of the two perspectives of input data can be scarce for some drugs or targets, especially for those unpopular or newly discovered. Furthermore, ground-truth labels for specific interaction types can also be scarce. Therefore, we propose the first method to tackle DTI prediction under input data and/or label scarcity. To make our model functional when only one perspective of input data is available, we design two separate experts to process intrinsic and extrinsic data respectively and fuse them adaptively according to different samples. Furthermore, to make the two perspectives complement each other and remedy label scarcity, two experts synergize with each other in a mutually supervised way to exploit the enormous unlabeled data. Extensive experiments on 3 real-world datasets under different extents of input data scarcity and/or label scarcity demonstrate our model outperforms states of the art significantly and steadily, with a maximum improvement of 53.53%. We also test our model without any data scarcity and it still outperforms current methods.