Yi Wu

Hongyu Wang, Shuming Ma, Li Dong et al. · microsoft-research

The increasing size of large language models has posed challenges for deployment and raised concerns about environmental impact due to high energy consumption. In this work, we introduce BitNet, a scalable and stable 1-bit Transformer architecture designed for large language models. Specifically, we introduce BitLinear as a drop-in replacement of the nn.Linear layer in order to train 1-bit weights from scratch. Experimental results on language modeling show that BitNet achieves competitive performance while substantially reducing memory footprint and energy consumption, compared to state-of-the-art 8-bit quantization methods and FP16 Transformer baselines. Furthermore, BitNet exhibits a scaling law akin to full-precision Transformers, suggesting its potential for effective scaling to even larger language models while maintaining efficiency and performance benefits.

Botian Xu, Feng Gao, Chao Yu et al. · tsinghua

In this work, we introduce OmniDrones, an efficient and flexible platform tailored for reinforcement learning in drone control, built on Nvidia's Omniverse Isaac Sim. It employs a bottom-up design approach that allows users to easily design and experiment with various application scenarios on top of GPU-parallelized simulations. It also offers a range of benchmark tasks, presenting challenges ranging from single-drone hovering to over-actuated system tracking. In summary, we propose an open-sourced drone simulation platform, equipped with an extensive suite of tools for drone learning. It includes 4 drone models, 5 sensor modalities, 4 control modes, over 10 benchmark tasks, and a selection of widely used RL baselines. To showcase the capabilities of OmniDrones and to support future research, we also provide preliminary results on these benchmark tasks. We hope this platform will encourage further studies on applying RL to practical drone systems.

Weihua Du, Jinglun Zhao, Chao Yu et al. · cmu, tsinghua

Truss layout design, namely finding a lightweight truss layout satisfying all the physical constraints, is a fundamental problem in the building industry. Generating the optimal layout is a challenging combinatorial optimization problem, which can be extremely expensive to solve by exhaustive search. Directly applying end-to-end reinforcement learning (RL) methods to truss layout design is infeasible either, since only a tiny portion of the entire layout space is valid under the physical constraints, leading to particularly sparse rewards for RL training. In this paper, we develop AutoTruss, a two-stage framework to efficiently generate both lightweight and valid truss layouts. AutoTruss first adopts Monte Carlo tree search to discover a diverse collection of valid layouts. Then RL is applied to iteratively refine the valid solutions. We conduct experiments and ablation studies in popular truss layout design test cases in both 2D and 3D settings. AutoTruss outperforms the best-reported layouts by 25.1% in the most challenging 3D test cases, resulting in the first effective deep-RL-based approach in the truss layout design literature.

Wenbin Zou, Tian Ye, Weixin Zheng et al.

Recently, deep learning has been successfully applied to the single-image super-resolution (SISR) with remarkable performance. However, most existing methods focus on building a more complex network with a large number of layers, which can entail heavy computational costs and memory storage. To address this problem, we present a lightweight Self-Calibrated Efficient Transformer (SCET) network to solve this problem. The architecture of SCET mainly consists of the self-calibrated module and efficient transformer block, where the self-calibrated module adopts the pixel attention mechanism to extract image features effectively. To further exploit the contextual information from features, we employ an efficient transformer to help the network obtain similar features over long distances and thus recover sufficient texture details. We provide comprehensive results on different settings of the overall network. Our proposed method achieves more remarkable performance than baseline methods. The source code and pre-trained models are available at https://github.com/AlexZou14/SCET.

Yunfei Li, Tao Kong, Lei Li et al. · bytedance, cmu

Can a robot autonomously learn to design and construct a bridge from varying-sized blocks without a blueprint? It is a challenging task with long horizon and sparse reward -- the robot has to figure out physically stable design schemes and feasible actions to manipulate and transport blocks. Due to diverse block sizes, the state space and action trajectories are vast to explore. In this paper, we propose a hierarchical approach for this problem. It consists of a reinforcement-learning designer to propose high-level building instructions and a motion-planning-based action generator to manipulate blocks at the low level. For high-level learning, we develop a novel technique, prioritized memory resetting (PMR) to improve exploration. PMR adaptively resets the state to those most critical configurations from a replay buffer so that the robot can resume training on partial architectures instead of from scratch. Furthermore, we augment PMR with auxiliary training objectives and fine-tune the designer with the locomotion generator. Our experiments in simulation and on a real deployed robotic system demonstrate that it is able to effectively construct bridges with blocks of varying sizes at a high success rate. Demos can be found at https://sites.google.com/view/bridge-pmr.

Jianwei Zhao, Feifei Gao, Qihui Wu et al.

Unmanned aerial vehicle (UAV)-satellite communication has drawn dramatic attention for its potential to build the integrated space-air-ground network and the seamless wide-area coverage. The key challenge to UAV-satellite communication is its unstable beam pointing due to the UAV navigation, which is a typical SatCom on-the-move scenario. In this paper, we propose a blind beam tracking approach for Ka-band UAVsatellite communication system, where UAV is equipped with a large-scale antenna array. The effects of UAV navigation are firstly released through the mechanical adjustment, which could approximately point the beam towards the target satellite through beam stabilization and dynamic isolation. Specially, the attitude information can be realtimely derived from data fusion of lowcost sensors. Then, the precision of the beam pointing is blindly refined through electrically adjusting the weight of the massive antennas, where an array structure based simultaneous perturbation algorithm is designed. Simulation results are provided to demonstrate the superiority of the proposed method over the existing ones.

Kan Xie, Zhe Zhang, Bo Li et al.

With the gradual popularization of self-driving, it is becoming increasingly important for vehicles to smartly make the right driving decisions and autonomously obey traffic rules by correctly recognizing traffic signs. However, for machine learning-based traffic sign recognition on the Internet of Vehicles (IoV), a large amount of traffic sign data from distributed vehicles is needed to be gathered in a centralized server for model training, which brings serious privacy leakage risk because of traffic sign data containing lots of location privacy information. To address this issue, we first exploit privacy-preserving federated learning to perform collaborative training for accurate recognition models without sharing raw traffic sign data. Nevertheless, due to the limited computing and energy resources of most devices, it is hard for vehicles to continuously undertake complex artificial intelligence tasks. Therefore, we introduce powerful Spike Neural Networks (SNNs) into traffic sign recognition for energy-efficient and fast model training, which is the next generation of neural networks and is practical and well-fitted to IoV scenarios. Furthermore, we design a novel encoding scheme for SNNs based on neuron receptive fields to extract information from the pixel and spatial dimensions of traffic signs to achieve high-accuracy training. Numerical results indicate that the proposed federated SNN outperforms traditional federated convolutional neural networks in terms of accuracy, noise immunity, and energy efficiency as well.

Liyuan Zhang, Jiarui Zhang, Jinwei Yao et al.

Speculative decoding accelerates autoregressive large language model inference by drafting multiple tokens and verifying them in a single target-model forward pass. Recent diffusion-based drafters generate an entire block of tokens in parallel but usually commit to a single draft sequence per verification: once the first mismatch occurs, all subsequent draft tokens are discarded, resulting in a limited acceptance rate. Naively batching more draft candidate sequences only introduces a marginal improvement, as redundant or poorly placed branches increase the cost of drafting and verification without proportionally increasing the number of accepted tokens. We propose D^2SD, a dual diffusion draft speculative decoding framework that organizes candidates into a confidence-guided prefix tree, where the first diffusion drafter generates a block along with per-position confidence scores that are used to identify the most likely rejection boundary and select the top-K prefix ranges for recovery; the second variable-prefix diffusion drafter re-anchors at each selected prefix and proposes alternative continuations in one batched pass; the resulting shared-prefix candidates are jointly verified via cascade attention. Empirically, D^2SD shows clear improvements over both the underlying diffusion approach and strong autoregressive speculative decoding baselines.

Yi Wu, Ziqiang Li, Chaoyue Wang et al.

In this study, we delve into the task of few-shot Generative Domain Adaptation (GDA), which involves transferring a pre-trained generator from one domain to a new domain using only a few reference images. Inspired by the way human brains acquire knowledge in new domains, we present an innovative generator structure called Domain Re-Modulation (DoRM). DoRM not only meets the criteria of high quality, large synthesis diversity, and cross-domain consistency, which were achieved by previous research in GDA, but also incorporates memory and domain association, akin to how human brains operate. Specifically, DoRM freezes the source generator and introduces new mapping and affine modules (M&A modules) to capture the attributes of the target domain during GDA. This process resembles the formation of new synapses in human brains. Consequently, a linearly combinable domain shift occurs in the style space. By incorporating multiple new M&A modules, the generator gains the capability to perform high-fidelity multi-domain and hybrid-domain generation. Moreover, to maintain cross-domain consistency more effectively, we introduce a similarity-based structure loss. This loss aligns the auto-correlation map of the target image with its corresponding auto-correlation map of the source image during training. Through extensive experiments, we demonstrate the superior performance of our DoRM and similarity-based structure loss in few-shot GDA, both quantitatively and qualitatively. The code will be available at https://github.com/wuyi2020/DoRM.

Chao Yu, Xinyi Yang, Jiaxuan Gao et al. · bytedance

We consider the problem of cooperative exploration where multiple robots need to cooperatively explore an unknown region as fast as possible. Multi-agent reinforcement learning (MARL) has recently become a trending paradigm for solving this challenge. However, existing MARL-based methods adopt action-making steps as the metric for exploration efficiency by assuming all the agents are acting in a fully synchronous manner: i.e., every single agent produces an action simultaneously and every single action is executed instantaneously at each time step. Despite its mathematical simplicity, such a synchronous MARL formulation can be problematic for real-world robotic applications. It can be typical that different robots may take slightly different wall-clock times to accomplish an atomic action or even periodically get lost due to hardware issues. Simply waiting for every robot being ready for the next action can be particularly time-inefficient. Therefore, we propose an asynchronous MARL solution, Asynchronous Coordination Explorer (ACE), to tackle this real-world challenge. We first extend a classical MARL algorithm, multi-agent PPO (MAPPO), to the asynchronous setting and additionally apply action-delay randomization to enforce the learned policy to generalize better to varying action delays in the real world. Moreover, each navigation agent is represented as a team-size-invariant CNN-based policy, which greatly benefits real-robot deployment by handling possible robot lost and allows bandwidth-efficient intra-agent communication through low-dimensional CNN features. We first validate our approach in a grid-based scenario. Both simulation and real-robot results show that ACE reduces over 10% actual exploration time compared with classical approaches. We also apply our framework to a high-fidelity visual-based environment, Habitat, achieving 28% improvement in exploration efficiency.

Ziqiang Li, Hong Sun, Pengfei Xia et al.

Recent deep neural networks (DNNs) have came to rely on vast amounts of training data, providing an opportunity for malicious attackers to exploit and contaminate the data to carry out backdoor attacks. However, existing backdoor attack methods make unrealistic assumptions, assuming that all training data comes from a single source and that attackers have full access to the training data. In this paper, we introduce a more realistic attack scenario where victims collect data from multiple sources, and attackers cannot access the complete training data. We refer to this scenario as data-constrained backdoor attacks. In such cases, previous attack methods suffer from severe efficiency degradation due to the entanglement between benign and poisoning features during the backdoor injection process. To tackle this problem, we introduce three CLIP-based technologies from two distinct streams: Clean Feature Suppression and Poisoning Feature Augmentation.effective solution for data-constrained backdoor attacks. The results demonstrate remarkable improvements, with some settings achieving over 100% improvement compared to existing attacks in data-constrained scenarios. Code is available at https://github.com/sunh1113/Efficient-backdoor-attacks-for-deep-neural-networks-in-real-world-scenarios

Zhiyu Mei, Wei Fu, Jiaxuan Gao et al.

The ever-growing complexity of reinforcement learning (RL) tasks demands a distributed system to efficiently generate and process a massive amount of data. However, existing open-source libraries suffer from various limitations, which impede their practical use in challenging scenarios where large-scale training is necessary. In this paper, we present a novel abstraction on the dataflows of RL training, which unifies diverse RL training applications into a general framework. Following this abstraction, we develop a scalable, efficient, and extensible distributed RL system called ReaLlyScalableRL, which allows efficient and massively parallelized training and easy development of customized algorithms. Our evaluation shows that SRL outperforms existing academic libraries, reaching at most 21x higher training throughput in a distributed setting. On learning performance, beyond performing and scaling well on common RL benchmarks with different RL algorithms, SRL can reproduce the same solution in the challenging hide-and-seek environment as reported by OpenAI with up to 5x speedup in wall-clock time. Notably, SRL is the first in the academic community to perform RL experiments at a large scale with over 15k CPU cores. SRL source code is available at: https://github.com/openpsi-project/srl .

Wangyi Mei, Zhouhong Gu, Zhenhan Bai et al.

Open-ended reasoning and long-form generation tasks lack reliable automatic verification signals for reward-based policy optimization. Rubrics offer a promising alternative, but existing approaches treat them as given artifacts -- either hand-crafted or prompt-generated -- and often miss the task-specific, knowledge-intensive dimensions that matter most, distorting the reward signal. Our key observation is that rubric construction is itself a research problem: identifying what makes a response correct or insightful requires discovering and synthesizing external knowledge. We propose Deep Research as Rubric (DR-rubric), a two-stage framework for constructing such rubrics. Stage I elicits domain facts, structural constraints, and failure modes through iterative multi-turn agentic search; Stage II distills this evidence into atomic, independently verifiable constraints for GRPO-based policy optimization. Because the model under training can serve as its own rubric generator, DR-rubric-8B supports bootstrap rubric generation without frontier-model assistance. We evaluate on 6 benchmarks spanning agentic research and expert reasoning. Experiments show that DR-Rubric achieves strong competitive performance with only 1K -- 3K training instances, where GPT-5-generated rubrics particularly benefit breadth coverage on agentic tasks, Gemini-generated rubrics yield the most balanced performance across agentic and expert reasoning tasks, and bootstrap rubrics exhibit a specialization-to-rebalancing evolution achieving the best overall performance at the third iteration. Results demonstrate that reframing rubric construction from static evaluation templates into an evidence-driven research process yields more scalable, fine-grained reward signals for open-ended tasks.

Lu Yang, Zelai Xu, Minyang Xie et al. · tsinghua

Large Language Model (LLM) agents have demonstrated remarkable proficiency in learned tasks, yet they often struggle to adapt to non-stationary environments with feedback. While In-Context Learning and external memory offer some flexibility, they fail to internalize the adaptive ability required for long-term improvement. Meta-Reinforcement Learning (meta-RL) provides an alternative by embedding the learning process directly within the model. However, existing meta-RL approaches for LLMs focus primarily on exploration in single-agent settings, neglecting the strategic exploitation necessary for multi-agent environments. We propose MAGE, a meta-RL framework that empowers LLM agents for strategic exploration and exploitation. MAGE utilizes a multi-episode training regime where interaction histories and reflections are integrated into the context window. By using the final episode reward as the objective, MAGE incentivizes the agent to refine its strategy based on past experiences. We further combine population-based training with an agent-specific advantage normalization technique to enrich agent diversity and ensure stable learning. Experiment results show that MAGE outperforms existing baselines in both exploration and exploitation tasks. Furthermore, MAGE exhibits strong generalization to unseen opponents, suggesting it has internalized the ability for strategic exploration and exploitation. Code is available at https://github.com/Lu-Yang666/MAGE.

Chao Yu, Jiaxuan Gao, Weilin Liu et al. · tsinghua

There is a recent trend of applying multi-agent reinforcement learning (MARL) to train an agent that can cooperate with humans in a zero-shot fashion without using any human data. The typical workflow is to first repeatedly run self-play (SP) to build a policy pool and then train the final adaptive policy against this pool. A crucial limitation of this framework is that every policy in the pool is optimized w.r.t. the environment reward function, which implicitly assumes that the testing partners of the adaptive policy will be precisely optimizing the same reward function as well. However, human objectives are often substantially biased according to their own preferences, which can differ greatly from the environment reward. We propose a more general framework, Hidden-Utility Self-Play (HSP), which explicitly models human biases as hidden reward functions in the self-play objective. By approximating the reward space as linear functions, HSP adopts an effective technique to generate an augmented policy pool with biased policies. We evaluate HSP on the Overcooked benchmark. Empirical results show that our HSP method produces higher rewards than baselines when cooperating with learned human models, manually scripted policies, and real humans. The HSP policy is also rated as the most assistive policy based on human feedback.

Wei Fu, Chao Yu, Zelai Xu et al. · tsinghua

Many advances in cooperative multi-agent reinforcement learning (MARL) are based on two common design principles: value decomposition and parameter sharing. A typical MARL algorithm of this fashion decomposes a centralized Q-function into local Q-networks with parameters shared across agents. Such an algorithmic paradigm enables centralized training and decentralized execution (CTDE) and leads to efficient learning in practice. Despite all the advantages, we revisit these two principles and show that in certain scenarios, e.g., environments with a highly multi-modal reward landscape, value decomposition, and parameter sharing can be problematic and lead to undesired outcomes. In contrast, policy gradient (PG) methods with individual policies provably converge to an optimal solution in these cases, which partially supports some recent empirical observations that PG can be effective in many MARL testbeds. Inspired by our theoretical analysis, we present practical suggestions on implementing multi-agent PG algorithms for either high rewards or diverse emergent behaviors and empirically validate our findings on a variety of domains, ranging from the simplified matrix and grid-world games to complex benchmarks such as StarCraft Multi-Agent Challenge and Google Research Football. We hope our insights could benefit the community towards developing more general and more powerful MARL algorithms. Check our project website at https://sites.google.com/view/revisiting-marl.

Huyu Wu, Jun Liu, Xiaochi Wei et al.

Self-evolving search agents reduce reliance on human-written training questions by generating and solving their own search tasks. We build on Search Self-Play (SSP), a representative Proposer and Solver framework in which questions are generated and answered via multi-step search and reasoning. In practice, however, SSP faces two bottlenecks: the Proposer constructs questions from isolated answer entities without relational context, yielding many invalid or unverifiable questions in early self-play training, while the Solver receives only a binary outcome reward that discards useful signal from partially on-track search trajectories. We address both bottlenecks by reusing knowledge-graph paths as construction-derived intermediate supervision for both question construction and reward shaping. First, we ground question construction in LLM-guided knowledge-graph subgraphs, providing relational context for the Proposer. Second, we observe that constructing and solving a multi-hop question can involve overlapping intermediate entities: the factual bridges used to formulate the question may provide approximate waypoints for answering it. Exploiting this overlap, we introduce Waypoint Coverage Reward (WCR), which grants graded partial credit to incorrect Solver trajectories according to their coverage of entities on the construction path, while preserving full reward for correct answers. Across seven QA benchmarks and nine model configurations, our approach improves the average score over standard SSP in all configurations, including notable gains on multi-hop QA tasks. These results suggest that knowledge-graph paths can be reused as lightweight intermediate supervision, providing both relational guidance and process feedback without additional task-specific human annotations or manually labeled process steps.

Yunfei Li, Tian Gao, Jiaqi Yang et al. · bytedance, tsinghua

It has been a recent trend to leverage the power of supervised learning (SL) towards more effective reinforcement learning (RL) methods. We propose a novel phasic approach by alternating online RL and offline SL for tackling sparse-reward goal-conditioned problems. In the online phase, we perform RL training and collect rollout data while in the offline phase, we perform SL on those successful trajectories from the dataset. To further improve sample efficiency, we adopt additional techniques in the online phase including task reduction to generate more feasible trajectories and a value-difference-based intrinsic reward to alleviate the sparse-reward issue. We call this overall algorithm, PhAsic self-Imitative Reduction (PAIR). PAIR substantially outperforms both non-phasic RL and phasic SL baselines on sparse-reward goal-conditioned robotic control problems, including a challenging stacking task. PAIR is the first RL method that learns to stack 6 cubes with only 0/1 success rewards from scratch.

Yi Wu, Zikang Xiong, Yiran Hu et al.

Despite significant advancements in large language models (LLMs) that enhance robot agents' understanding and execution of natural language (NL) commands, ensuring the agents adhere to user-specified constraints remains challenging, particularly for complex commands and long-horizon tasks. To address this challenge, we present three key insights, equivalence voting, constrained decoding, and domain-specific fine-tuning, which significantly enhance LLM planners' capability in handling complex tasks. Equivalence voting ensures consistency by generating and sampling multiple Linear Temporal Logic (LTL) formulas from NL commands, grouping equivalent LTL formulas, and selecting the majority group of formulas as the final LTL formula. Constrained decoding then uses the generated LTL formula to enforce the autoregressive inference of plans, ensuring the generated plans conform to the LTL. Domain-specific fine-tuning customizes LLMs to produce safe and efficient plans within specific task domains. Our approach, Safe Efficient LLM Planner (SELP), combines these insights to create LLM planners to generate plans adhering to user commands with high confidence. We demonstrate the effectiveness and generalizability of SELP across different robot agents and tasks, including drone navigation and robot manipulation. For drone navigation tasks, SELP outperforms state-of-the-art planners by 10.8% in safety rate (i.e., finishing tasks conforming to NL commands) and by 19.8% in plan efficiency. For robot manipulation tasks, SELP achieves 20.4% improvement in safety rate. Our datasets for evaluating NL-to-LTL and robot task planning will be released in github.com/lt-asset/selp.

Zhecheng Yuan, Zhengrong Xue, Bo Yuan et al.

Learning generalizable policies that can adapt to unseen environments remains challenging in visual Reinforcement Learning (RL). Existing approaches try to acquire a robust representation via diversifying the appearances of in-domain observations for better generalization. Limited by the specific observations of the environment, these methods ignore the possibility of exploring diverse real-world image datasets. In this paper, we investigate how a visual RL agent would benefit from the off-the-shelf visual representations. Surprisingly, we find that the early layers in an ImageNet pre-trained ResNet model could provide rather generalizable representations for visual RL. Hence, we propose Pre-trained Image Encoder for Generalizable visual reinforcement learning (PIE-G), a simple yet effective framework that can generalize to the unseen visual scenarios in a zero-shot manner. Extensive experiments are conducted on DMControl Generalization Benchmark, DMControl Manipulation Tasks, Drawer World, and CARLA to verify the effectiveness of PIE-G. Empirical evidence suggests PIE-G improves sample efficiency and significantly outperforms previous state-of-the-art methods in terms of generalization performance. In particular, PIE-G boasts a 55% generalization performance gain on average in the challenging video background setting. Project Page: https://sites.google.com/view/pie-g/home.

Xin Hu, Zhenyu Wu, Hao-Yu Miao et al.

Video action detection (spatio-temporal action localization) is usually the starting point for human-centric intelligent analysis of videos nowadays. It has high practical impacts for many applications across robotics, security, healthcare, etc. The two-stage paradigm of Faster R-CNN inspires a standard paradigm of video action detection in object detection, i.e., firstly generating person proposals and then classifying their actions. However, none of the existing solutions could provide fine-grained action detection to the "who-when-where-what" level. This paper presents a tracking-based solution to accurately and efficiently localize predefined key actions spatially (by predicting the associated target IDs and locations) and temporally (by predicting the time in exact frame indices). This solution won first place in the UAV-Video Track of 2021 Low-Power Computer Vision Challenge (LPCVC).

Xuyang Zhi, Peilun zhou, Chengqiang Lu et al.

The evolution of Large Language Models (LLMs) is shifting the focus from single, verifiable tasks toward complex, open-ended real-world scenarios, imposing significant challenges on the post-training phase. In these settings, the scale and complexity of reward systems have grown significantly, transitioning toward multi-objective formulations that encompass a comprehensive spectrum of model capabilities and application contexts. However, traditional methods typically rely on fixed reward weights, ignoring non-stationary learning dynamics and struggling with data heterogeneity across dimensions. To address these issues, we propose SPARD, a framework that establishes an automated, self-paced curriculum by perceiving learning progress to dynamically adjust multi-objective reward weights and data importance, thereby synchronizing learning intent with data utility for optimal performance. Extensive experiments across multiple benchmarks demonstrate that SPARD significantly enhances model capabilities across all domains.

Jiayu Chen, Zelai Xu, Yunfei Li et al. · bytedance

Learning Nash equilibrium (NE) in complex zero-sum games with multi-agent reinforcement learning (MARL) can be extremely computationally expensive. Curriculum learning is an effective way to accelerate learning, but an under-explored dimension for generating a curriculum is the difficulty-to-learn of the subgames -- games induced by starting from a specific state. In this work, we present a novel subgame curriculum learning framework for zero-sum games. It adopts an adaptive initial state distribution by resetting agents to some previously visited states where they can quickly learn to improve performance. Building upon this framework, we derive a subgame selection metric that approximates the squared distance to NE values and further adopt a particle-based state sampler for subgame generation. Integrating these techniques leads to our new algorithm, Subgame Automatic Curriculum Learning (SACL), which is a realization of the subgame curriculum learning framework. SACL can be combined with any MARL algorithm such as MAPPO. Experiments in the particle-world environment and Google Research Football environment show SACL produces much stronger policies than baselines. In the challenging hide-and-seek quadrant environment, SACL produces all four emergent stages and uses only half the samples of MAPPO with self-play. The project website is at https://sites.google.com/view/sacl-rl.

Wei Fu, Weihua Du, Jingwei Li et al. · cmu

In complex reinforcement learning (RL) problems, policies with similar rewards may have substantially different behaviors. It remains a fundamental challenge to optimize rewards while also discovering as many diverse strategies as possible, which can be crucial in many practical applications. Our study examines two design choices for tackling this challenge, i.e., diversity measure and computation framework. First, we find that with existing diversity measures, visually indistinguishable policies can still yield high diversity scores. To accurately capture the behavioral difference, we propose to incorporate the state-space distance information into the diversity measure. In addition, we examine two common computation frameworks for this problem, i.e., population-based training (PBT) and iterative learning (ITR). We show that although PBT is the precise problem formulation, ITR can achieve comparable diversity scores with higher computation efficiency, leading to improved solution quality in practice. Based on our analysis, we further combine ITR with two tractable realizations of the state-distance-based diversity measures and develop a novel diversity-driven RL algorithm, State-based Intrinsic-reward Policy Optimization (SIPO), with provable convergence properties. We empirically examine SIPO across three domains from robot locomotion to multi-agent games. In all of our testing environments, SIPO consistently produces strategically diverse and human-interpretable policies that cannot be discovered by existing baselines.

Shangfei Wang, Yanan Chang, Yi Wu et al.

Facial affective behavior analysis is important for human-computer interaction. 5th ABAW competition includes three challenges from Aff-Wild2 database. Three common facial affective analysis tasks are involved, i.e. valence-arousal estimation, expression classification, action unit recognition. For the three challenges, we construct three different models to solve the corresponding problems to improve the results, such as data unbalance and data noise. For the experiments of three challenges, we train the models on the provided training data and validate the models on the validation data.

Zikun Li, Jinjun Peng, Yixuan Mei et al.

Optimizing quantum circuits is challenging due to the very large search space of functionally equivalent circuits and the necessity of applying transformations that temporarily decrease performance to achieve a final performance improvement. This paper presents Quarl, a learning-based quantum circuit optimizer. Applying reinforcement learning (RL) to quantum circuit optimization raises two main challenges: the large and varying action space and the non-uniform state representation. Quarl addresses these issues with a novel neural architecture and RL-training procedure. Our neural architecture decomposes the action space into two parts and leverages graph neural networks in its state representation, both of which are guided by the intuition that optimization decisions can be mostly guided by local reasoning while allowing global circuit-wide reasoning. Our evaluation shows that Quarl significantly outperforms existing circuit optimizers on almost all benchmark circuits. Surprisingly, Quarl can learn to perform rotation merging, a complex, non-local circuit optimization implemented as a separate pass in existing optimizers.

Jingyang Ou, Jiaqi Han, Minkai Xu et al.

Reinforcement Learning (RL) has proven highly effective for autoregressive language models, but adapting these methods to diffusion large language models (dLLMs) presents fundamental challenges. The core difficulty lies in likelihood approximation: while autoregressive models naturally provide token-level conditional probabilities essential for token-level RL objectives (e.g., GRPO), dLLMs generate sequences through iterative non-autoregressive denoising steps that lack this factorization. To address this fundamental mismatch, we propose ELBO-based Sequence-level Policy Optimization (ESPO), a principled RL framework that treats entire sequence generation as a single action and uses the ELBO as a tractable sequence-level likelihood proxy. Our method incorporates per-token normalization of importance ratios and robust KL-divergence estimation to ensure stable large-scale training. Extensive experiments on mathematical reasoning, coding, and planning tasks demonstrate that ESPO significantly outperforms token-level baselines, achieving dramatic improvements of 20-40 points on the Countdown task, while maintaining consistent gains on math and coding benchmarks. Our approach establishes sequence-level optimization as a principled and empirically effective paradigm for RL in dLLMs. Our code is available at https://github.com/ML-GSAI/ESPO.

Zelai Xu, Chao Yu, Fei Fang et al.

Agents built with large language models (LLMs) have shown great potential across a wide range of domains. However, in complex decision-making tasks, pure LLM-based agents tend to exhibit intrinsic bias in their choice of actions, which is inherited from the model's training data and results in suboptimal performance. To develop strategic language agents, i.e., agents that generate flexible language actions and possess strong decision-making abilities, we propose a novel framework that powers LLM-based agents with reinforcement learning (RL). We consider Werewolf, a popular social deduction game, as a challenging testbed that emphasizes versatile communication and strategic gameplay. To mitigate the intrinsic bias in language actions, our agents use an LLM to perform deductive reasoning and generate a diverse set of action candidates. Then an RL policy trained to optimize the decision-making ability chooses an action from the candidates to play in the game. Extensive experiments show that our agents overcome the intrinsic bias and outperform existing LLM-based agents in the Werewolf game. We also conduct human-agent experiments and find that our agents achieve human-level performance and demonstrate strong strategic play.

Yanan Chang, Yi Wu, Xiangyu Miao et al.

Facial valence/arousal, expression and action unit are related tasks in facial affective analysis. However, the tasks only have limited performance in the wild due to the various collected conditions. The 4th competition on affective behavior analysis in the wild (ABAW) provided images with valence/arousal, expression and action unit labels. In this paper, we introduce multi-task learning framework to enhance the performance of three related tasks in the wild. Feature sharing and label fusion are used to utilize their relations. We conduct experiments on the provided training and validating data.

Jiaxuan Gao, Jiaao Chen, Chuyi He et al. · tsinghua

Interactive tool-using agents must solve real-world tasks via multi-turn interaction with both humans and external environments, requiring dialogue state tracking, multi-step tool execution, while following complex instructions. Post-training such agents is challenging because synthesis for high-quality multi-turn tool-use data is difficult to scale, and reinforcement learning (RL) could face noisy signals caused by user simulation, leading to degraded training efficiency. We propose a unified framework that combines a self-evolving data agent with verifier-based RL. Our system, EigenData, is a hierarchical multi-agent engine that synthesizes tool-grounded dialogues together with executable per-instance checkers, and improves generation reliability via closed-loop self-evolving process that updates prompts and workflow. Building on the synthetic data, we develop an RL recipe that first fine-tunes the user model and then applies GRPO-style training with trajectory-level group-relative advantages and dynamic filtering, yielding consistent improvements beyond SFT. Evaluated on tau^2-bench, our best model reaches 73.0% pass^1 on Airline and 98.3% pass^1 on Telecom, matching or exceeding frontier models. Overall, our results suggest a scalable pathway for bootstrapping complex tool-using behaviors without expensive human annotation.

Zixuan Yang, Yiqun Chen, Wei Yang et al.

Reinforcement learning in open-ended long-form generation is challenging because reliable reference answers and automatic metrics are often unavailable. Existing rubric-based methods typically rely on pointwise LLM-as-a-judge scoring, but absolute scores are difficult to calibrate across complex responses, may provide weak discrimination among same-query rollouts, and can become saturated during optimization. We propose Tournament-GRPO, a group-wise reward framework that converts rubric-guided LLM judgments into relative rewards through repeated multi-round tournaments among same-query rollouts. Tournament-GRPO compares candidates within groups, accumulates tournament outcomes, and normalizes them into group-wise rewards for GRPO training. Experiments on Deep Research Bench show that Tournament-GRPO consistently outperforms existing reward-design baselines, achieving a 4.52-point overall-score improvement over the strongest baseline. Further analyses show that tournament rewards provide a favorable effectiveness--efficiency trade-off and that tournament design affects training dynamics. These results suggest that rubric-guided tournament comparison provides an effective reward signal for reinforcement learning in open-ended long-form generation.

Yiqun Chen, Wei Yang, Erhan Zhang et al.

LLM-based multi-agent systems decompose complex tasks into interacting roles, but most remain manually orchestrated by prompts, tools, and control rules, while agents are rarely optimized through a unified reinforcement learning interface. Existing RL post-training frameworks mainly target single-policy optimization and lack abstractions for user-defined multi-agent workflows, structured interaction, role-specific credit assignment, and configurable parameter sharing. We present UnityMAS-O, a general RL optimization framework for LLM-based multi-agent systems. UnityMAS-O treats the complete workflow as the optimization unit, rather than a single response or policy trajectory. It represents workflows through four first-class objects: logical agent roles, graph trajectories, user-defined rewards, and agent--model mappings. This decouples logical agents from physical model parameters, supporting full sharing, full separation, and partial sharing, with rewards assigned at role, turn, and trajectory levels. UnityMAS-O extends verl with a Ray-based star-topology runtime. A central controller executes workflows, invokes tools, records structured trajectories, and assembles rewards; model-local worker groups handle rollout, buffering, advantage computation, and distributed PPO-style updates. Users can define agents, workflows, model mappings, and rewards without rewriting the optimization infrastructure. We instantiate UnityMAS-O on retrieval-augmented QA, iterative agentic search, and reflective code generation. Across Natural Questions, HotpotQA, and held-out code tasks, multi-agent RL improves manually specified workflows after optimization, with especially large gains for smaller models and strict code all-passed metrics. These results show that UnityMAS-O can serve as a reusable substrate for converting diverse LLM-based multi-agent workflows into trainable multi-agent RL systems.

Jing Wang, Meichen Song, Feng Gao et al.

We initiate the study of how to perturb the reward in a zero-sum Markov game with two players to induce a desirable Nash equilibrium, namely arbitrating. Such a problem admits a bi-level optimization formulation. The lower level requires solving the Nash equilibrium under a given reward function, which makes the overall problem challenging to optimize in an end-to-end way. We propose a backpropagation scheme that differentiates through the Nash equilibrium, which provides the gradient feedback for the upper level. In particular, our method only requires a black-box solver for the (regularized) Nash equilibrium (NE). We develop the convergence analysis for the proposed framework with proper black-box NE solvers and demonstrate the empirical successes in two multi-agent reinforcement learning (MARL) environments.

Zihan Zhou, Wei Fu, Bingliang Zhang et al.

We present Reward-Switching Policy Optimization (RSPO), a paradigm to discover diverse strategies in complex RL environments by iteratively finding novel policies that are both locally optimal and sufficiently different from existing ones. To encourage the learning policy to consistently converge towards a previously undiscovered local optimum, RSPO switches between extrinsic and intrinsic rewards via a trajectory-based novelty measurement during the optimization process. When a sampled trajectory is sufficiently distinct, RSPO performs standard policy optimization with extrinsic rewards. For trajectories with high likelihood under existing policies, RSPO utilizes an intrinsic diversity reward to promote exploration. Experiments show that RSPO is able to discover a wide spectrum of strategies in a variety of domains, ranging from single-agent particle-world tasks and MuJoCo continuous control to multi-agent stag-hunt games and StarCraftII challenges.

Shusheng Xu, Wei Fu, Jiaxuan Gao et al. · tsinghua

Reinforcement Learning from Human Feedback (RLHF) is currently the most widely used method to align large language models (LLMs) with human preferences. Existing RLHF methods can be roughly categorized as either reward-based or reward-free. Novel applications such as ChatGPT and Claude leverage reward-based methods that first learn a reward model and apply actor-critic algorithms, such as Proximal Policy Optimization (PPO). However, in academic benchmarks, state-of-the-art results are often achieved via reward-free methods, such as Direct Preference Optimization (DPO). Is DPO truly superior to PPO? Why does PPO perform poorly on these benchmarks? In this paper, we first conduct both theoretical and empirical studies on the algorithmic properties of DPO and show that DPO may have fundamental limitations. Moreover, we also comprehensively examine PPO and reveal the key factors for the best performances of PPO in fine-tuning LLMs. Finally, we benchmark DPO and PPO across a collection of RLHF testbeds, ranging from dialogue to code generation. Experiment results demonstrate that PPO is able to surpass other alignment methods in all cases and achieve state-of-the-art results in challenging code competitions. Our code is publicly available at https://github.com/openpsi-project/ReaLHF.

Shangfei Wang, Jiaqiang Wu, Feiyi Zheng et al.

This paper introduces our method for the Emotional Reaction Intensity (ERI) Estimation Challenge, in CVPR 2023: 5th Workshop and Competition on Affective Behavior Analysis in-the-wild (ABAW). Based on the multimodal data provided by the originazers, we extract acoustic and visual features with different pretrained models. The multimodal features are mixed together by Transformer Encoders with cross-modal attention mechnism. In this paper, 1. better features are extracted with the SOTA pretrained models. 2. Compared with the baseline, we improve the Pearson's Correlations Coefficient a lot. 3. We process the data with some special skills to enhance performance ability of our model.

Yun-Shiuan Chuang, Yi Wu, Dhruv Gupta et al.

Adapting pre-trained language models (PLMs) for time-series text classification amidst evolving domain shifts (EDS) is critical for maintaining accuracy in applications like stance detection. This study benchmarks the effectiveness of evolving domain adaptation (EDA) strategies, notably self-training, domain-adversarial training, and domain-adaptive pretraining, with a focus on an incremental self-training method. Our analysis across various datasets reveals that this incremental method excels at adapting PLMs to EDS, outperforming traditional domain adaptation techniques. These findings highlight the importance of continually updating PLMs to ensure their effectiveness in real-world applications, paving the way for future research into PLM robustness against the natural temporal evolution of language.

Xiangyu Miao, Jiahe Wang, Yanan Chang et al.

Learning from synthetic images plays an important role in facial expression recognition task due to the difficulties of labeling the real images, and it is challenging because of the gap between the synthetic images and real images. The fourth Affective Behavior Analysis in-the-wild Competition raises the challenge and provides the synthetic images generated from Aff-Wild2 dataset. In this paper, we propose a hand-assisted expression recognition method to reduce the gap between the synthetic data and real data. Our method consists of two parts: expression recognition module and hand prediction module. Expression recognition module extracts expression information and hand prediction module predicts whether the image contains hands. Decision mode is used to combine the results of two modules, and post-pruning is used to improve the result. F1 score is used to verify the effectiveness of our method.

Xincheng Yao, Ruoqi Li, Cheng Chen et al.

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a pivotal technique for enhancing the reasoning capabilities of Large Language Models (LLMs). However, the de facto practice of mainstream RL algorithms is to treat all tokens of one response equally and assign the same optimization objective to each token, failing to provide granular guidance for the reasoning process. While in Chain-of-Thought (CoT) reasoning, different tokens usually play distinct roles. Therefore, the current RL algorithms lack an effective mechanism to dynamically balance the exploration-exploitation trade-off during learning. To this end, we propose Hierarchical Token-level Objective Control Policy Optimization (HTPO), a novel RL algorithm that takes the divide-and-conquer idea to hierarchically partition the response tokens into specific functional groups from three aspects (i.e., prompt difficulty, answer correctness, and token entropy). Within each group, according to the contributions to exploration or exploitation, we design specialized optimization objectives to facilitate the effective execution of each token's expected functionality. In this way, HTPO can achieve a more balanced exploration-exploitation trade-off. Extensive experiments on challenging reasoning benchmarks validate the superiority of our HTPO algorithm, which significantly outperforms the strong DAPO baseline (e.g., +8.6% and +6.7% on AIME'24 and AIME'25, respectively). When scaling test-time compute, the HTPO-trained model maintains a consistent performance advantage over the DAPO baseline, and the gap widens as the sampling budget increases, validating that our adaptive token-level control method fosters effective exploration without sacrificing exploitation performance. Code will be at https://github.com/xcyao00/HTPO.

Zelai Xu, Yancheng Liang, Chao Yu et al.

Self-play (SP) is a popular multi-agent reinforcement learning (MARL) framework for solving competitive games, where each agent optimizes policy by treating others as part of the environment. Despite the empirical successes, the theoretical properties of SP-based methods are limited to two-player zero-sum games. However, for mixed cooperative-competitive games where agents on the same team need to cooperate with each other, we can show a simple counter-example where SP-based methods cannot converge to a global Nash equilibrium (NE) with high probability. Alternatively, Policy-Space Response Oracles (PSRO) is an iterative framework for learning NE, where the best responses w.r.t. previous policies are learned in each iteration. PSRO can be directly extended to mixed cooperative-competitive settings by jointly learning team best responses with all convergence properties unchanged. However, PSRO requires repeatedly training joint policies from scratch till convergence, which makes it hard to scale to complex games. In this work, we develop a novel algorithm, Fictitious Cross-Play (FXP), which inherits the benefits from both frameworks. FXP simultaneously trains an SP-based main policy and a counter population of best response policies. The main policy is trained by fictitious self-play and cross-play against the counter population, while the counter policies are trained as the best responses to the main policy's past versions. We validate our method in matrix games and show that FXP converges to global NEs while SP methods fail. We also conduct experiments in a gridworld domain, where FXP achieves higher Elo ratings and lower exploitabilities than baselines, and a more challenging football game, where FXP defeats SOTA models with over 94% win rate.

Wei Fu, Jiaxuan Gao, Xujie Shen et al. · tsinghua

Reinforcement learning (RL) has become a dominant paradigm for training large language models (LLMs), particularly for reasoning tasks. Effective RL for LLMs requires massive parallelization and poses an urgent need for efficient training systems. Most existing large-scale RL systems for LLMs are synchronous, alternating generation and training in a batch setting where rollouts in each training batch are generated by the same model. This approach stabilizes RL training but suffers from severe system-level inefficiency: generation must wait until the longest output in the batch is completed before model updates, resulting in GPU underutilization. We present AReaL, a fully asynchronous RL system that completely decouples generation from training. Rollout workers in AReaL continuously generate new outputs without waiting, while training workers update the model whenever a batch of data is collected. AReaL also incorporates a collection of system-level optimizations, leading to substantially higher GPU utilization. To stabilize RL training, AReaL balances the workload of rollout and training workers to control data staleness, and adopts a staleness-enhanced PPO variant to better handle outdated training samples. Extensive experiments on math and code reasoning benchmarks show that AReaL achieves up to 2.77$\times$ training speedup compared to synchronous systems with the same number of GPUs and matched or improved final performance. The code of AReaL is available at https://github.com/inclusionAI/AReaL/.

Yancheng Liang, Jiajie Zhang, Hui Li et al.

Despite the tremendous advances achieved over the past years by deep learning techniques, the latest risk prediction models for industrial applications still rely on highly handtuned stage-wised statistical learning tools, such as gradient boosting and random forest methods. Different from images or languages, real-world financial data are high-dimensional, sparse, noisy and extremely imbalanced, which makes deep neural network models particularly challenging to train and fragile in practice. In this work, we propose DeRisk, an effective deep learning risk prediction framework for credit risk prediction on real-world financial data. DeRisk is the first deep risk prediction model that outperforms statistical learning approaches deployed in our company's production system. We also perform extensive ablation studies on our method to present the most critical factors for the empirical success of DeRisk.

Pei Li, Zhijun Liu, Luyi Chang et al.

With the development and progress of science and technology, the Internet of Things(IoT) has gradually entered people's lives, bringing great convenience to our lives and improving people's work efficiency. Specifically, the IoT can replace humans in jobs that they cannot perform. As a new type of IoT vehicle, the current status and trend of research on Unmanned Aerial Vehicle(UAV) is gratifying, and the development prospect is very promising. However, privacy and communication are still very serious issues in drone applications. This is because most drones still use centralized cloud-based data processing, which may lead to leakage of data collected by drones. At the same time, the large amount of data collected by drones may incur greater communication overhead when transferred to the cloud. Federated learning as a means of privacy protection can effectively solve the above two problems. However, federated learning when applied to UAV networks also needs to consider the heterogeneity of data, which is caused by regional differences in UAV regulation. In response, this paper proposes a new algorithm FedBA to optimize the global model and solves the data heterogeneity problem. In addition, we apply the algorithm to some real datasets, and the experimental results show that the algorithm outperforms other algorithms and improves the accuracy of the local model for UAVs.

Jiaxuan Gao, Wei Fu, Minyang Xie et al. · tsinghua

Recent advancements in LLM-based agents have demonstrated remarkable capabilities in handling complex, knowledge-intensive tasks by integrating external tools. Among diverse choices of tools, search tools play a pivotal role in accessing vast external knowledge. However, open-source agents still fall short of achieving expert-level Search Intelligence, the ability to resolve ambiguous queries, generate precise searches, analyze results, and conduct thorough exploration. Existing approaches fall short in scalability, efficiency, and data quality. For example, small turn limits in existing online RL methods, e.g. <=10, restrict complex strategy learning. This paper introduces ASearcher, an open-source project for large-scale RL training of search agents. Our key contributions include: (1) Scalable fully asynchronous RL training that enables long-horizon search while maintaining high training efficiency. (2) A prompt-based LLM agent that autonomously synthesizes high-quality and challenging QAs, creating a large-scale QA dataset. Through RL training, our prompt-based QwQ-32B agent achieves substantial improvements, with 78.0% and 34.3% Avg@4 gains on xBench and GAIA, respectively. Notably, our agent exhibits extreme long-horizon search, with tool calls exceeding 100 turns and output tokens exceeding 400k during training time. With a simple agent design and no external LLMs, ASearcher-Web-QwQ achieves Avg@4 scores of 51.1 on xBench and 58.7 on GAIA, surpassing existing open-source 32B agents. Finally, we also show that ASearcher-Web-QwQ could achieve performance of commercial systems using external summary tool in a zero-shot transfer manner and test-time search. We open-source our models, training data, and codes in https://github.com/inclusionAI/ASearcher.

Guanqi Zhan, Changye Li, Zhijian Liu et al.

Visual grounding is an essential capability of Visual Language Models (VLMs) to understand the real physical world. Previous state-of-the-art grounding visual language models usually have large model sizes, making them heavy for deployment and slow for inference. However, we notice that the sizes of visual encoders are nearly the same for small and large VLMs and the major difference is the sizes of the language models. Small VLMs fall behind larger VLMs in grounding because of the difference in language understanding capability rather than visual information handling. To mitigate the gap, we introduce 'Efficient visual Grounding language Models' (EGM): a method to scale the test-time computation (#generated tokens). Scaling the test-time computation of a small model is deployment-friendly, and yields better end-to-end latency as the cost of each token is much cheaper compared to directly running a large model. On the RefCOCO benchmark, our EGM-Qwen3-VL-8B demonstrates 91.4 IoU with an average of 737ms (5.9x faster) latency while Qwen3-VL-235B demands 4,320ms to achieve 90.5 IoU. To validate our approach's generality, we further set up a new amodal grounding setting that requires the model to predict both the visible and occluded parts of the objects. Experiments show our method can consistently and significantly improve the vanilla grounding and amodal grounding capabilities of small models to be on par with or outperform the larger models, thereby improving the efficiency for visual grounding.

Jinyuan Feng, Xin Yu, Yiqun Chen et al.

The iterative denoising paradigm of Diffusion Large Language Models (DLMs) endows them with a distinct advantage in global context modeling. However, current decoding strategies fail to leverage this capability, typically exhibiting a local preference that overlooks the heterogeneous information density within the context, ultimately degrading generation quality. To address this limitation, we systematically investigate high-information-density (HD) tokens and present two key findings: (1) explicitly conditioning on HD tokens substantially improves output quality; and (2) HD tokens exhibit an early-decoding tendency, converging earlier than surrounding tokens. Motivated by these findings, we propose Focus on the Core \textbf{(FoCore)}, a training-free decoding strategy that utilizes HD tokens in a self-contrast manner, wherein HD tokens are temporarily remasked as negative samples, to guide generation. We further introduce FoCore\_Accelerate \textbf{(FoCore\_A)}, an efficient variant that, upon detecting HD token convergence, performs parallel decoding over stable candidates within a local context window, substantially accelerating generation. Extensive experiments on math, code and logical reasoning benchmarks demonstrate that FoCore consistently improves generation quality and efficiency across both LLaDA and Dream backbones. For instance, on HumanEval, FoCore improves pass@1 from 39.02 to 42.68 over standard Classifier-Free Guidance, while FoCore-A reduces the number of decoding steps by 2.07x and per-sample latency from 20.76s to 8.64s (-58.4\%).

Jiayu Chen, Chao Yu, Yuqing Xie et al. · tsinghua

Executing precise and agile flight maneuvers is critical for quadrotors in various applications. Traditional quadrotor control approaches are limited by their reliance on flat trajectories or time-consuming optimization, which restricts their flexibility. Recently, RL-based policy has emerged as a promising alternative due to its ability to directly map observations to actions, reducing the need for detailed system knowledge and actuation constraints. However, a significant challenge remains in bridging the sim-to-real gap, where RL-based policies often experience instability when deployed in real world. In this paper, we investigate key factors for learning robust RL-based control policies that are capable of zero-shot deployment in real-world quadrotors. We identify five critical factors and we develop a PPO-based training framework named SimpleFlight, which integrates these five techniques. We validate the efficacy of SimpleFlight on Crazyflie quadrotor, demonstrating that it achieves more than a 50% reduction in trajectory tracking error compared to state-of-the-art RL baselines. The policy derived by SimpleFlight consistently excels across both smooth polynominal trajectories and challenging infeasible zigzag trajectories on small thrust-to-weight quadrotors. In contrast, baseline methods struggle with high-speed or infeasible trajectories. To support further research and reproducibility, we integrate SimpleFlight into a GPU-based simulator Omnidrones and provide open-source access to the code and model checkpoints. We hope SimpleFlight will offer valuable insights for advancing RL-based quadrotor control. For more details, visit our project website at https://sites.google.com/view/simpleflight/.

Yi Wu, Daryl Chang, Jennifer She et al.

We present the Learned Ranking Function (LRF), a system that takes short-term user-item behavior predictions as input and outputs a slate of recommendations that directly optimizes for long-term user satisfaction. Most previous work is based on optimizing the hyperparameters of a heuristic function. We propose to model the problem directly as a slate optimization problem with the objective of maximizing long-term user satisfaction. We also develop a novel constraint optimization algorithm that stabilizes objective trade-offs for multi-objective optimization. We evaluate our approach with live experiments and describe its deployment on YouTube.

Jiazheng Li, Hongzhou Lin, Hong Lu et al.

Reinforcement learning (RL) has emerged as a central paradigm for training large language models (LLMs) in reasoning tasks. Yet recent studies question RL's ability to incentivize reasoning capacity beyond the base model. This raises a key challenge: how can RL be adapted to solve harder reasoning problems more effectively? To address this challenge, we propose a simple yet effective strategy via Question Augmentation: introduce partial solutions during training to reduce problem difficulty and provide more informative learning signals. Our method, QuestA, when applied during RL training on math reasoning tasks, not only improves pass@1 but also pass@k-particularly on problems where standard RL struggles to make progress. This enables continual improvement over strong open-source models such as DeepScaleR and OpenMath Nemotron, further enhancing their reasoning capabilities. We achieve new state-of-the-art results on math benchmarks using 1.5B-parameter models: 72.50% (+10.73%) on AIME24, 62.29% (+12.79%) on AIME25, and 41.67% (+10.11%) on HMMT25. Code, data and model are available at https://github.com/foreverlasting1202/QuestA.

Feng Gao, Chao Yu, Yu Wang et al.

Accurate motion control in the face of disturbances within complex environments remains a major challenge in robotics. Classical model-based approaches often struggle with nonlinearities and unstructured disturbances, while RL-based methods can be fragile when encountering unseen scenarios. In this paper, we propose a novel framework, Neural Internal Model Control, which integrates model-based control with RL-based control to enhance robustness. Our framework streamlines the predictive model by applying Newton-Euler equations for rigid-body dynamics, eliminating the need to capture complex high-dimensional nonlinearities. This internal model combines model-free RL algorithms with predictive error feedback. Such a design enables a closed-loop control structure to enhance the robustness and generalizability of the control system. We demonstrate the effectiveness of our framework on both quadrotors and quadrupedal robots, achieving superior performance compared to state-of-the-art methods. Furthermore, real-world deployment on a quadrotor with rope-suspended payloads highlights the framework's robustness in sim-to-real transfer. Our code is released at https://github.com/thu-uav/NeuralIMC.