Tao Ren

Tao Ren, Weiyao Luo, Hui Yang et al.

Existing memory-augmented LLM agents store past experience exclusively in prompt space, as textual summaries or retrieved passages, while keeping model parameters frozen throughout a rollout. Such agents can \emph{look up} what they have seen but cannot \emph{learn from} it: their policy is unchanged by experience, and any information dropped from the context is permanently lost. We introduce \texttt{TMEM}, a self-evolving parametric memory framework in which the agent not only compresses history into explicit memory but also absorbs distilled supervision into fast LoRA weights $Δ_t$ via lightweight online updates, genuinely altering its future behavior within a single episode. We formalize this as an agentic decision process with fast-weight rollout dynamics: actions are sampled from $π_{θ_0+Δ_t}$, while extraction actions produce supervision that updates $Δ_t$ for subsequent decisions. This view makes the extraction policy directly optimizable by RL: training $θ_0$ improves not only task actions but also the quality of the data used for online LoRA adaptation. We further propose SVD-based initialization of the LoRA subspace to accelerate online convergence. Experiments on LoCoMo, LongMemEval-S, multi-objective search, and CL-Bench show that \texttt{TMEM} consistently outperforms summary-based and retrieval-based baselines across different model scales.

Zexin Li, Tao Ren, Xiaoxi He et al.

Intelligent robots are designed to effectively navigate dynamic and unpredictable environments laden with moving mechanical elements and objects. Such environment-induced dynamics, including moving obstacles, can readily alter the computational demand (e.g., the creation of new tasks) and the structure of workloads (e.g., precedence constraints among tasks) during runtime, thereby adversely affecting overall system performance. This challenge is amplified when multi-task inference is expected on robots operating under stringent resource and real-time constraints. To address such a challenge, we introduce RED, a systematic real-time scheduling approach designed to support multi-task deep neural network workloads in resource-limited robotic systems. It is designed to adaptively manage the Robotic Environmental Dynamics (RED) while adhering to real-time constraints. At the core of RED lies a deadline-based scheduler that employs an intermediate deadline assignment policy, effectively managing to change workloads and asynchronous inference prompted by complex, unpredictable environments. This scheduling framework also facilitates the flexible deployment of MIMONet (multi-input multi-output neural networks), which are commonly utilized in multi-tasking robotic systems to circumvent memory bottlenecks. Building on this scheduling framework, RED recognizes and leverages a unique characteristic of MIMONet: its weight-shared architecture. To further accommodate and exploit this feature, RED devises a novel and effective workload refinement and reconstruction process. This process ensures the scheduling framework's compatibility with MIMONet and maximizes efficiency.

Zexin Li, Tao Ren, Johnathan Liu et al.

Robots deployed in dynamic environments must contend with environment-driven changes that reshape computation at runtime: new tasks may appear, precedence relations can shift, and overall workload structure evolves, all of which degrade performance, especially when multi-task inference is required under tight resource and real-time budgets. We present RED, a real-time scheduling framework for multi-task deep neural network workloads on resource-constrained robotic platforms that adapts to Robotic Environmental Dynamics (RED) while preserving end-to-end timing guarantees under modeling assumptions. The core of RED is a deadline-aware scheduler that assigns intermediate sub-deadlines, allowing it to accommodate evolving computation graphs and asynchronous inference induced by unpredictable conditions. The framework also supports flexible deployment of MIMONet (multi-input multi-output neural networks), commonly used in multi-tasking robots to alleviate memory pressure through weight sharing. RED explicitly leverages this shared-parameter property via a workload refinement and graph-reconstruction procedure that aligns MIMONet structure with schedulability requirements, improving compatibility and efficiency. We implement RED on NVIDIA Jetson family platforms and on an Apple M-series MacBook and evaluate it on navigation-oriented workloads representative of real robotic scenarios. Experiments show consistent gains over existing methods in throughput, deadline satisfaction, robustness to interference, adaptability, and runtime overhead.

Yiming Zeng, Wanhao Yu, Zexin Li et al.

Large Language Models (LLMs) have significantly advanced natural language processing, demonstrating strong capabilities in tasks such as text generation, summarization, and reasoning. Recently, their potential for automating precise text editing tasks across specialized domains, such as programming code, LaTeX, and structured database languages, has gained attention. However, current state-of-the-art LLMs still struggle with executing precise, instruction-driven edits, particularly when structural accuracy and strict adherence to domain conventions are required. To address these challenges, we introduce InstrEditBench, an automated benchmark dataset comprising over 30,000 structured editing tasks spanning diverse domains, including Wikipedia articles, LaTeX documents, source code, and database languages. Using this benchmark, we develop FineEdit, a specialized editing model explicitly trained for accurate, context-aware text modifications. Experimental evaluations demonstrate that FineEdit outperforms state-of-the-art models, achieving improvements of approximately 10\% over Gemini models on single-turn edits, up to 30\% over Llama-3.2-3B, and exceeding Mistral-7B-OpenOrca performance by over 40\% on direct editing tasks. FineEdit also effectively generalizes to realistic multi-turn editing scenarios, highlighting its practical applicability. To facilitate further research and reproducibility, we release FineEdit at https://github.com/StuRinDQB/FineEdit} and https://huggingface.co/datasets/YimingZeng/FineEdit_bench.

Jiafeng Liang, Hao Li, Chang Li et al.

Memory serves as the pivotal nexus bridging past and future, providing both humans and AI systems with invaluable concepts and experience to navigate complex tasks. Recent research on autonomous agents has increasingly focused on designing efficient memory workflows by drawing on cognitive neuroscience. However, constrained by interdisciplinary barriers, existing works struggle to assimilate the essence of human memory mechanisms. To bridge this gap, we systematically synthesizes interdisciplinary knowledge of memory, connecting insights from cognitive neuroscience with LLM-driven agents. Specifically, we first elucidate the definition and function of memory along a progressive trajectory from cognitive neuroscience through LLMs to agents. We then provide a comparative analysis of memory taxonomy, storage mechanisms, and the complete management lifecycle from both biological and artificial perspectives. Subsequently, we review the mainstream benchmarks for evaluating agent memory. Additionally, we explore memory security from dual perspectives of attack and defense. Finally, we envision future research directions, with a focus on multimodal memory systems and skill acquisition.

Zhongyi Li, Wan Tian, Jingyu Chen et al.

Multi-agent collaboration has emerged as a powerful paradigm for enhancing the reasoning capabilities of large language models, yet it suffers from interaction-level ambiguity that blurs generation, critique, and revision, making credit assignment across agents difficult. Moreover, policy optimization in this setting is vulnerable to heavy-tailed and noisy rewards, which can bias advantage estimation and trigger unstable or even divergent training. To address both issues, we propose a robust multi-agent reinforcement learning framework for collaborative reasoning, consisting of two components: Dual-Agent Answer-Critique-Rewrite (DACR) and an Adaptive Robust Estimator (ARE). DACR decomposes reasoning into a structured three-stage pipeline: answer, critique, and rewrite, while enabling explicit attribution of each agent's marginal contribution to its partner's performance. ARE provides robust estimation of batch experience means during multi-agent policy optimization. Across mathematical reasoning and embodied intelligence benchmarks, even under noisy rewards, our method consistently outperforms the baseline in both homogeneous and heterogeneous settings. These results indicate stronger robustness to reward noise and more stable training dynamics, effectively preventing optimization failures caused by noisy reward signals.

Jiafeng Liang, Zhihao Zhu, Zihan Zhang et al.

Although Large Multimodal Models (LMMs) have achieved strong performance on general video understanding, their susceptibility to textual prior shortcuts during causal discovery has been recognized as a critical deficit. The underlying mechanisms of this phenomenon remain incompletely understood, as existing benchmarks only measure response accuracy without revealing the sources and extent of the deficit. We introduce ProCauEval, a perturbation-based evaluation protocol that shifts from outcome assessment to mechanism diagnosis, probing causal discovery through five controlled configurations that systematically manipulate visual and textual modalities to decompose their respective contributions to model behavior and dissect the failure modes. Evaluating 17 mainstream LMMs, we find that models faithfully perceive video content yet systematically underexploit it during causal reasoning. We further observe that stronger post-training amplifies rather than mitigates textual prior reliance, and that higher baseline performance correlates with greater fragility under perturbation. To address these, we propose Anti-Distillation Policy Optimization (ADPO), a reinforcement learning framework built on negative teacher alignment, which augments GRPO by explicitly pushing the policy away from a prior-only counterfactual teacher induced by visual corruption. Specifically, ADPO maximizes the divergence between the policy distributions conditioned on the original and visually corrupted inputs, thereby forcing the model to ground its reasoning in visual evidence rather than textual shortcuts. Extensive experiments show that ADPO improves visual engagement without sacrificing fundamental comprehension, thus offering a preliminary step toward reliable causal discovery.

Zehao Li, Tao Ren, Zishi Zhang et al.

Large language model (LLM) training is often bottlenecked by memory constraints and stochastic gradient noise in extremely high-dimensional parameter spaces. Motivated by empirical evidence that many LLM gradient matrices are effectively low-rank during training, we present an unbiased, memory-efficient, low-rank matrix estimator with the lowest variance that is applicable across common stochastic gradient estimation paradigms. The core idea is to project a high-dimensional stochastic gradient estimator onto a random low-dimensional subspace and lift it back, reducing memory while keeping the estimator unbiased and controlling mean-squared error via an optimally designed projection distribution, including Haar--Stiefel projections. The projection distribution is derived by solving a constrained functional optimization problem, yielding an optimal random projector that guides algorithm design. Empirically, the resulting low-rank gradient estimators deliver both practical memory savings and improved training behavior. In RoBERTa-large fine-tuning, our method attains the lowest peak GPU memory among compared methods (e.g., 3.83GB versus 16.7GB for full BP) while remaining competitive in accuracy; in autoregressive LLM pretraining (LLaMA-20M/60M/100M), our method outperforms the traditional methods, supporting the benefit of the proposed optimal projection strategy.

Tao Ren, Xiaoyu Luo, Qiongxiu Li

Prior work on probing neural networks primarily relies on input-space analysis or parameter perturbation, both of which face fundamental limitations in accessing structural information encoded in intermediate representations. We introduce Activation Perturbation for EXploration (APEX), an inference-time probing paradigm that perturbs hidden activations while keeping both inputs and model parameters fixed. We theoretically show that activation perturbation induces a principled transition from sample-dependent to model-dependent behavior by suppressing input-specific signals and amplifying representation-level structure, and further establish that input perturbation corresponds to a constrained special case of this framework. Through representative case studies, we demonstrate the practical advantages of APEX. In the small-noise regime, APEX provides a lightweight and efficient measure of sample regularity that aligns with established metrics, while also distinguishing structured from randomly labeled models and revealing semantically coherent prediction transitions. In the large-noise regime, APEX exposes training-induced model-level biases, including a pronounced concentration of predictions on the target class in backdoored models. Overall, our results show that APEX offers an effective perspective for exploring, and understanding neural networks beyond what is accessible from input space alone.

Ziheng Xu, Jianwei Niu, Qingfeng Li et al.

Neural implicit representations have been explored to enhance visual SLAM algorithms, especially in providing high-fidelity dense map. Existing methods operate robustly in static scenes but struggle with the disruption caused by moving objects. In this paper we present NID-SLAM, which significantly improves the performance of neural SLAM in dynamic environments. We propose a new approach to enhance inaccurate regions in semantic masks, particularly in marginal areas. Utilizing the geometric information present in depth images, this method enables accurate removal of dynamic objects, thereby reducing the probability of camera drift. Additionally, we introduce a keyframe selection strategy for dynamic scenes, which enhances camera tracking robustness against large-scale objects and improves the efficiency of mapping. Experiments on publicly available RGB-D datasets demonstrate that our method outperforms competitive neural SLAM approaches in tracking accuracy and mapping quality in dynamic environments.

Hui Yang, Tao Ren, Jinyang Jiang et al.

Memory-efficient optimization methods have recently gained increasing attention for scaling full-parameter training of large language models under the GPU-memory bottleneck. Existing approaches either lack clear convergence guarantees, or only achieve the standard ${\mathcal{O}}(ε^{-4})$ iteration complexity in the nonconvex settings. We propose Omni-Masked Gradient Descent (OMGD), an optimization method based on mask traversal for memory efficient training, and provide a nonconvex convergence analysis that establishes a strictly improved iteration complexity of $\tilde{\mathcal{O}}(ε^{-3})$ for finding an $ε$-approximate stationary point. Empirically, OMGD is a lightweight, plug-and-play approach that integrates seamlessly into most mainstream optimizers, yielding consistent improvements over competitive baselines in both fine-tuning and pre-training tasks.

Bin Lei, Yuchen Li, Yiming Zeng et al.

Despite the impressive capabilities of large language models (LLMs), they currently exhibit two primary limitations, \textbf{\uppercase\expandafter{\romannumeral 1}}: They struggle to \textbf{autonomously solve the real world engineering problem}. \textbf{\uppercase\expandafter{\romannumeral 2}}: They remain \textbf{challenged in reasoning through complex logic problems}. To address these challenges, we developed the \textsc{Infant Agent}, integrating task-aware functions, operators, a hierarchical management system, and a memory retrieval mechanism. Together, these components enable large language models to sustain extended reasoning processes and handle complex, multi-step tasks efficiently, all while significantly reducing API costs. Using the \textsc{Infant Agent}, GPT-4o's accuracy on the SWE-bench-lite dataset rises from $\mathbf{0.33\%}$ to $\mathbf{30\%}$, and in the AIME-2024 mathematics competition, it increases GPT-4o's accuracy from $\mathbf{13.3\%}$ to $\mathbf{37\%}$.

Yiming Zeng, Jinghan Cao, Zexin Li et al.

Recent advances in diffusion-based language models have opened new possibilities for controllable and bidirectional sequence generation. These models provide an alternative to traditional autoregressive approaches by framing text generation as an iterative denoising process. However, applying diffusion models to structured domains such as source code remains a significant challenge. Programming languages differ from natural language in that they follow strict syntactic and semantic rules, with hierarchical organization that must be preserved for correctness. Standard token-level corruption techniques used during training often ignore this structure, which may hinder the model's ability to learn meaningful representations of code. To address this limitation, we propose a syntax-aware diffusion framework that incorporates structural priors from Abstract Syntax Trees (ASTs) into the denoising process. Instead of masking individual tokens at random, we selectively corrupt syntactically meaningful code spans derived from AST subtrees. This enables the model to reconstruct programs in a way that respects grammatical boundaries and captures long-range dependencies. Experimental results demonstrate that syntax-aware corruption significantly improves syntactic correctness, reconstruction accuracy, and generalization to unseen code patterns. These findings highlight the potential of incorporating structural information into diffusion-based training and suggest that syntax-guided denoising is a promising direction for advancing diffusion-based language models in code generation tasks.

Tao Ren, Zishi Zhang, Jingyang Jiang et al.

The probabilistic diffusion model (DM), generating content by inferencing through a recursive chain structure, has emerged as a powerful framework for visual generation. After pre-training on enormous data, the model needs to be properly aligned to meet requirements for downstream applications. How to efficiently align the foundation DM is a crucial task. Contemporary methods are either based on Reinforcement Learning (RL) or truncated Backpropagation (BP). However, RL and truncated BP suffer from low sample efficiency and biased gradient estimation, respectively, resulting in limited improvement or, even worse, complete training failure. To overcome the challenges, we propose the Recursive Likelihood Ratio (RLR) optimizer, a Half-Order (HO) fine-tuning paradigm for DM. The HO gradient estimator enables the computation graph rearrangement within the recursive diffusive chain, making the RLR's gradient estimator an unbiased one with lower variance than other methods. We theoretically investigate the bias, variance, and convergence of our method. Extensive experiments are conducted on image and video generation to validate the superiority of the RLR. Furthermore, we propose a novel prompt technique that is natural for the RLR to achieve a synergistic effect.

Tao Ren, Jinyang Jiang, Hui Yang et al. · pku

Reinforcement learning with verifiable reward has recently emerged as a central paradigm for post-training large language models (LLMs); however, prevailing mean-based methods, such as Group Relative Policy Optimization (GRPO), suffer from entropy collapse and limited reasoning gains. We argue that these issues stem from overemphasizing high-probability output sequences while neglecting rare but informative reasoning paths. To address these challenges, we propose Risk-based Policy Optimization (RiskPO), which substitutes classical mean-based objectives with principled risk measures. Specifically, we introduce a Mixed Value-at-Risk objective that integrates weighted attention over multiple regions of the reward distribution, thereby amplifying gradient signals on challenging instances and preventing overconfident convergence. We further design a bundling scheme that aggregates multiple questions into bundles, thus enriching the feedback signal and yielding more stable and informative training dynamics. Theoretically, we prove that the risk-averse update alleviates entropy collapse and promotes exploration. Numerically, RiskPO achieves consistent and significant improvements in mathematical reasoning, multi-modal reasoning, and code generation benchmarks, surpassing GRPO and its variants on both Pass@1 and Pass@k metrics. Our results demonstrate that risk-based optimization provides a rigorous and effective paradigm for enhancing LLM reasoning capabilities.

Guanghao Li, Wenhao Jiang, Mingfeng Chen et al.

Chain of Thought (CoT) prompting improves the reasoning performance of large language models (LLMs) by encouraging step by step thinking. However, CoT-based methods depend on intermediate reasoning steps, which limits scalability and generalization. Recent work explores recursive reasoning, where LLMs reuse internal layers across iterations to refine latent representations without explicit CoT supervision. While promising, these approaches often require costly pretraining and lack a principled framework for how reasoning should evolve across iterations. We address this gap by introducing Flow Chain of Thought (Flow CoT), a reasoning paradigm that models recursive inference as a progressive trajectory of latent cognitive states. Flow CoT frames each iteration as a distinct cognitive stage deepening reasoning across iterations without relying on manual supervision. To realize this, we propose SCOUT (Stepwise Cognitive Optimization Using Teachers), a lightweight fine tuning framework that enables Flow CoT style reasoning without the need for pretraining. SCOUT uses progressive distillation to align each iteration with a teacher of appropriate capacity, and a cross attention based retrospective module that integrates outputs from previous iterations while preserving the models original computation flow. Experiments across eight reasoning benchmarks show that SCOUT consistently improves both accuracy and explanation quality, achieving up to 1.8% gains under fine tuning. Qualitative analyses further reveal that SCOUT enables progressively deeper reasoning across iterations refining both belief formation and explanation granularity. These results not only validate the effectiveness of SCOUT, but also demonstrate the practical viability of Flow CoT as a scalable framework for enhancing reasoning in LLMs.

Jinyang Jiang, Xiaotian Liu, Tao Ren et al. · pku

We introduce a deep reinforcement learning (DRL) approach for solving management problems including inventory management, dynamic pricing, and recommendation. This DRL approach has the potential to lead to a large management model based on certain transformer neural network structures, resulting in an artificial general intelligence paradigm for various management tasks. Traditional methods have limitations for solving complex real-world problems, and we demonstrate how DRL can surpass existing heuristic approaches for solving management tasks. We aim to solve the problems in a unified framework, considering the interconnections between different tasks. Central to our methodology is the development of a foundational decision model coordinating decisions across the different domains through generative decision-making. Our experimental results affirm the effectiveness of our DRL-based framework in complex and dynamic business environments. This work opens new pathways for the application of DRL in management problems, highlighting its potential to revolutionize traditional business management.

Hao Su, Jianwei Niu, Xuefeng Liu et al.

Quick Response (QR) code is one of the most worldwide used two-dimensional codes.~Traditional QR codes appear as random collections of black-and-white modules that lack visual semantics and aesthetic elements, which inspires the recent works to beautify the appearances of QR codes. However, these works adopt fixed generation algorithms and therefore can only generate QR codes with a pre-defined style. In this paper, combining the Neural Style Transfer technique, we propose a novel end-to-end method, named ArtCoder, to generate the stylized QR codes that are personalized, diverse, attractive, and scanning-robust.~To guarantee that the generated stylized QR codes are still scanning-robust, we propose a Sampling-Simulation layer, a module-based code loss, and a competition mechanism. The experimental results show that our stylized QR codes have high-quality in both the visual effect and the scanning-robustness, and they are able to support the real-world application.

Guowei Zhang, Tao Ren, Yifan Yang

A new variational mode decomposition (VMD) based deep learning approach is proposed in this paper for time series forecasting problem. Firstly, VMD is adopted to decompose the original time series into several sub-signals. Then, a convolutional neural network (CNN) is applied to learn the reconstruction patterns on the decomposed sub-signals to obtain several reconstructed sub-signals. Finally, a long short term memory (LSTM) network is employed to forecast the time series with the decomposed sub-signals and the reconstructed sub-signals as inputs. The proposed VMD-CNN-LSTM approach is originated from the decomposition-reconstruction-ensemble framework, and innovated by embedding the reconstruction, single forecasting, and ensemble steps in a unified deep learning approach. To verify the forecasting performance of the proposed approach, four typical time series datasets are introduced for empirical analysis. The empirical results demonstrate that the proposed approach outperforms consistently the benchmark approaches in terms of forecasting accuracy, and also indicate that the reconstructed sub-signals obtained by CNN is of importance for further improving the forecasting performance.