Ting Long

Wei Liu, Ting Long

Cross-domain offline reinforcement learning (CDRL) aims to improve policy learning in a target domain by leveraging data collected from a source domain. Existing works typically assess the transferability of source-domain data by measuring its similarity to target-domain transitions, and implicitly perform transition-level selection. Transitions that are considered similar are assigned higher weights or rewards, while dissimilar ones are down-weighted. However, transition-level similarity does not necessarily imply consistency in long-term returns. Even visually or dynamically similar transitions may lead to significantly different outcomes in the target domain, which can mislead policy learning and degrade performance. To address this issue, we revisit the fundamental objective of policy learning. Since policy optimization ultimately relies on Bellman targets to evaluate the quality of decisions, we propose to assess the transferability of source-domain transitions based on their alignment with target-domain Bellman targets, rather than superficial transition similarity. Based on this insight, we propose a method termed Target-Aligned Bellman Backup (TABB), which selectively leverages source-domain data by measuring their contribution to accurate Bellman target estimation in the target domain. We evaluate TABB across a broad range of cross-domain offline RL settings with highly limited target-domain data. Experimental results show that TABB consistently achieves strong performance.

Yixiang Shan, Zhengbang Zhu, Ting Long et al.

The performance of offline reinforcement learning (RL) is sensitive to the proportion of high-return trajectories in the offline dataset. However, in many simulation environments and real-world scenarios, there are large ratios of low-return trajectories rather than high-return trajectories, which makes learning an efficient policy challenging. In this paper, we propose a method called Contrastive Diffuser (CDiffuser) to make full use of low-return trajectories and improve the performance of offline RL algorithms. Specifically, CDiffuser groups the states of trajectories in the offline dataset into high-return states and low-return states and treats them as positive and negative samples correspondingly. Then, it designs a contrastive mechanism to pull the trajectory of an agent toward high-return states and push them away from low-return states. Through the contrast mechanism, trajectories with low returns can serve as negative examples for policy learning, guiding the agent to avoid areas associated with low returns and achieve better performance. Experiments on 14 commonly used D4RL benchmarks demonstrate the effectiveness of our proposed method. Our code is publicly available at \url{https://anonymous.4open.science/r/CDiffuser}.

Huan Song, Shuyu Tian, Qingfei Zhao et al.

We present Ruyi2.5, a multimodal familial model built on the AI Flow framework. Extending Ruyi2's "Train Once, Deploy Many" paradigm to the multimodal domain, Ruyi2.5 constructs a shared-backbone architecture that co-trains models of varying scales within a single unified pipeline, ensuring semantic consistency across all deployment tiers. Built upon Ruyi2.5, Ruyi2.5-Camera model is developed as a privacy-preserving camera service system, which instantiates Ruyi2.5-Camera into a two-stage recognition pipeline: an edge model applies information-bottleneck-guided irreversible feature mapping to de-identify raw frames at the source, while a cloud model performs deep behavior reasoning. To accelerate reinforcement learning fine-tuning, we further propose Binary Prefix Policy Optimization (BPPO), which reduces sample redundancy via binary response selection and focuses gradient updates on response prefixes, achieving a 2 to 3 times training speedup over GRPO. Experiments show Ruyi2.5 matches Qwen3-VL on the general multimodal benchmarks, while Ruyi2.5-Camera substantially outperforms Qwen3-VL on privacy-constrained surveillance tasks.

Huan Song, Qingfei Zhao, Ting Long et al.

Neural scaling laws have become foundational for optimizing large language model (LLM) training, yet they typically assume a single dense model output. This limitation effectively overlooks "Familial models, a transformative paradigm essential for realizing ubiquitous intelligence across heterogeneous device-edge-cloud hierarchies. Transcending static architectures, familial models integrate early exits with relay-style inference to spawn G deployable sub-models from a single shared backbone. In this work, we theoretically and empirically extend the scaling law to capture this "one-run, many-models" paradigm by introducing Granularity (G) as a fundamental scaling variable alongside model size (N) and training tokens (D). To rigorously quantify this relationship, we propose a unified functional form L(N, D, G) and parameterize it using large-scale empirical runs. Specifically, we employ a rigorous IsoFLOP experimental design to strictly isolate architectural impact from computational scale. Across fixed budgets, we systematically sweep model sizes (N) and granularities (G) while dynamically adjusting tokens (D). This approach effectively decouples the marginal cost of granularity from the benefits of scale, ensuring high-fidelity parameterization of our unified scaling law. Our results reveal that the granularity penalty follows a multiplicative power law with an extremely small exponent. Theoretically, this bridges fixed-compute training with dynamic architectures. Practically, it validates the "train once, deploy many" paradigm, demonstrating that deployment flexibility is achievable without compromising the compute-optimality of dense baselines.

Haipeng Liu, Ting Long, Jing Fu

The question-answering (QA) simulator is a model that mimics real student learning behaviors and predicts their correctness of their responses to questions. QA simulators enable educational recommender systems (ERS) to collect large amounts of training data without interacting with real students, thereby preventing harmful recommendations made by an undertrained ERS from undermining actual student learning. Given the QA history, there are two categories of solutions to predict the correctness, conducting the simulation: (1) LLM-free methods, which apply a traditional sequential model to transfer the QA history into a vector representation first, and make predictions based on the representation; (2) LLM-based methods, which leverage the domain knowledge and reasoning capability of LLM to enhence the prediction. LLM-free methods offer fast inference but generally yield suboptimal performance. In contrast, most LLM-based methods achieve better results, but at the cost of slower inference speed and higher GPU memory consumption. In this paper, we propose a method named LLM Distillation based Simulator (LDSim), which distills domain knowledge and reasoning capability from an LLM to better assist prediction, thereby improving simulation performance. Extensive experiments demonstrate that our LDSim achieves strong results on both the simulation task and the knowledge tracing (KT) task. Our code is publicly available at https://anonymous.4open.science/r/LDSim-05A9.

Haipeng Liu, Yuxuan Liu, Ting Long

Personalized question recommendation aims to guide individual students through questions to enhance their mastery of learning targets. Most previous methods model this task as a Markov Decision Process and use reinforcement learning to solve, but they struggle with efficient exploration, failing to identify the best questions for each student during training. To address this, we propose Ranking Alignment Recommendation (RAR), which incorporates collaborative ideas into the exploration mechanism, enabling more efficient exploration within limited training episodes. Experiments show that RAR effectively improves recommendation performance, and our framework can be applied to any RL-based question recommender. Our code is available in https://github.com/wuming29/RAR.git.

Guanghe Li, Yixiang Shan, Zhengbang Zhu et al.

In offline reinforcement learning (RL), the performance of the learned policy highly depends on the quality of offline datasets. However, in many cases, the offline dataset contains very limited optimal trajectories, which poses a challenge for offline RL algorithms as agents must acquire the ability to transit to high-reward regions. To address this issue, we introduce Diffusion-based Trajectory Stitching (DiffStitch), a novel diffusion-based data augmentation pipeline that systematically generates stitching transitions between trajectories. DiffStitch effectively connects low-reward trajectories with high-reward trajectories, forming globally optimal trajectories to address the challenges faced by offline RL algorithms. Empirical experiments conducted on D4RL datasets demonstrate the effectiveness of DiffStitch across RL methodologies. Notably, DiffStitch demonstrates substantial enhancements in the performance of one-step methods (IQL), imitation learning methods (TD3+BC), and trajectory optimization methods (DT).

Lichang Song, Ting Long, Yi Chang

Retrieval-Augmented Generation (RAG) has demonstrated strong effectiveness in knowledge-intensive tasks by grounding language generation in external evidence. Despite its success, many existing RAG systems are built based on a ranking-centric, asymmetric dependency paradigm, where the generation quality of the generator is highly dependent on reranking results of the reranker. To overcome this limitation, we reformulate RAG as a cooperative multi-agent decision-making problem and propose Cooperative Retrieval-Augmented Generation (CoRAG), a framework in which the reranker and the generator act as peer decision-makers rather than being connected through an asymmetric dependency pipeline. By jointly optimizing their behaviors toward a shared task objective, the reranker and generator are encouraged to cooperate, ensuring that document reranking and generation work in concert to improve the final response. Experimental results demonstrate good generalization and improved generation stability of CoRAG, even when the model is trained on only around 10K PopQA samples. Our model released in https://anonymous.4open.science/r/CoRAG-D63F

Huan Song, Shuyu Tian, Ting Long et al.

With the increasing deployment of intelligent sensing technologies in highly sensitive environments such as restrooms and locker rooms, visual surveillance systems face a profound privacy-security paradox. Existing privacy-preserving approaches, including physical desensitization, encryption, and obfuscation, often compromise semantic understanding or fail to ensure mathematically provable irreversibility. Although Privacy Camera 1.0 eliminated visual data at the source to prevent leakage, it provided only textual judgments, leading to evidentiary blind spots in disputes. To address these limitations, this paper proposes a novel privacy-preserving perception framework based on the AI Flow paradigm and a collaborative edge-cloud architecture. By deploying a visual desensitizer at the edge, raw images are transformed in real time into abstract feature vectors through nonlinear mapping and stochastic noise injection under the Information Bottleneck principle, ensuring identity-sensitive information is stripped and original images are mathematically unreconstructable. The abstract representations are transmitted to the cloud for behavior recognition and semantic reconstruction via a "dynamic contour" visual language, achieving a critical balance between perception and privacy while enabling illustrative visual reference without exposing raw images.

Lu Guo, Yixiang Shan, Zhengbang Zhu et al.

Offline reinforcement learning (RL) enables agents to learn policies from fixed datasets, avoiding costly or unsafe environment interactions. However, its effectiveness is often limited by dataset sparsity and the lack of transition overlap between suboptimal and expert trajectories, which makes long-horizon planning particularly challenging. Prior solutions based on synthetic data augmentation or trajectory stitching often fail to generalize to novel states and rely on heuristic stitching points. To address these challenges, we propose Retrieval High-quAlity Demonstrations (RAD) for decision-making, which combines non-parametric retrieval with diffusion-based generative modeling. RAD dynamically retrieves high-return states from the offline dataset as target states based on state similarity and return estimation, and plans toward them using a condition-guided diffusion model. Such retrieval-guided generation enables flexible trajectory stitching and improves generalization when encountered with underrepresented or out-of-distribution states. Extensive experiments confirm that RAD achieves competitive or superior performance compared to baselines across diverse benchmarks, validating its effectiveness.

Lingyue Fu, Ting Long, Jianghao Lin et al.

Knowledge Tracing (KT) monitors students' knowledge states and simulates their responses to question sequences. Existing KT models typically follow a single-step training paradigm, which leads to discrepancies with the multi-step inference process required in real-world simulations, resulting in significant error accumulation. This accumulation of error, coupled with the issue of data sparsity, can substantially degrade the performance of recommendation models in the intelligent tutoring systems. To address these challenges, we propose a novel Adversarial Multi-Step Training Framework for Knowledge Tracing (AdvKT), which, for the first time, focuses on the multi-step KT task. More specifically, AdvKT leverages adversarial learning paradigm involving a generator and a discriminator. The generator mimics high-reward responses, effectively reducing error accumulation across multiple steps, while the discriminator provides feedback to generate synthetic data. Additionally, we design specialized data augmentation techniques to enrich the training data with realistic variations, ensuring that the model generalizes well even in scenarios with sparse data. Experiments conducted on four real-world datasets demonstrate the superiority of AdvKT over existing KT models, showcasing its ability to address both error accumulation and data sparsity issues effectively.

Ting Long, Yutong Xie, Xianyu Chen et al.

Program representation, which aims at converting program source code into vectors with automatically extracted features, is a fundamental problem in programming language processing (PLP). Recent work tries to represent programs with neural networks based on source code structures. However, such methods often focus on the syntax and consider only one single perspective of programs, limiting the representation power of models. This paper proposes a multi-view graph (MVG) program representation method. MVG pays more attention to code semantics and simultaneously includes both data flow and control flow as multiple views. These views are then combined and processed by a graph neural network (GNN) to obtain a comprehensive program representation that covers various aspects. We thoroughly evaluate our proposed MVG approach in the context of algorithm detection, an important and challenging subfield of PLP. Specifically, we use a public dataset POJ-104 and also construct a new challenging dataset ALG-109 to test our method. In experiments, MVG outperforms previous methods significantly, demonstrating our model's strong capability of representing source code.