Learning Query-Specific Rubrics from Human Preferences for DeepResearch Report Generation

Nowadays, training and evaluating DeepResearch-generated reports remain challenging due to the lack of verifiable reward signals. Accordingly, rubric-based evaluation has become a common practice. However, existing approaches either rely on coarse, pre-defined rubrics that lack sufficient granularity, or depend on manually constructed query-specific rubrics that are costly and difficult to scale. In this paper, we propose a pipeline to train human-preference-aligned query-specific rubric generators tailored for DeepResearch report generation. We first construct a dataset of DeepResearch-style queries annotated with human preferences over paired reports, and train rubric generators via reinforcement learning with a hybrid reward combining human preference supervision and LLM-based rubric evaluation. To better handle long-horizon reasoning, we further introduce a Multi-agent Markov-state (MaMs) workflow for report generation. We empirically show that our proposed rubric generators deliver more discriminative and better human-aligned supervision than existing rubric design strategies. Moreover, when integrated into the MaMs training framework, DeepResearch systems equipped with our rubric generators consistently outperform all open-source baselines on the DeepResearch Bench and achieve performance comparable to that of leading closed-source models.

Spatiotemporal Decision Transformer for Traffic Coordination

Traffic signal control is a critical challenge in urban transportation, requiring coordination among multiple intersections to optimize network-wide traffic flow. While reinforcement learning has shown promise for adaptive signal control, existing methods struggle with multi-agent coordination and sample efficiency. We introduce MADT (Multi-Agent Decision Transformer), a novel approach that reformulates multi-agent traffic signal control as a sequence modeling problem. MADT extends the Decision Transformer paradigm to multi-agent settings by incorporating: (1) a graph attention mechanism for modeling spatial dependencies between intersections, (2) a|temporal transformer encoder for capturing traffic dynamics, and (3) return-to-go conditioning for target performance specification. Our approach enables offline learning from historical traffic data, with architecture design that facilitates potential online fine-tuning. Experiments on synthetic grid networks and real-world traffic scenarios demonstrate that MADT achieves state-of-the-art performance, reducing average travel time by 5-6% compared to the strongest baseline while exhibiting superior coordination among adjacent intersections.

Explicit Credit Assignment through Local Rewards and Dependence Graphs in Multi-Agent Reinforcement Learning

To promote cooperation in Multi-Agent Reinforcement Learning, the reward signals of all agents can be aggregated together, forming global rewards that are commonly known as the fully cooperative setting. However, global rewards are usually noisy because they contain the contributions of all agents, which have to be resolved in the credit assignment process. On the other hand, using local reward benefits from faster learning due to the separation of agents' contributions, but can be suboptimal as agents myopically optimize their own reward while disregarding the global optimality. In this work, we propose a method that combines the merits of both approaches. By using a graph of interaction between agents, our method discerns the individual agent contribution in a more fine-grained manner than a global reward, while alleviating the cooperation problem with agents' local reward. We also introduce a practical approach for approximating such a graph. Our experiments demonstrate the flexibility of the approach, enabling improvements over the traditional local and global reward settings.

From Intents to Actions: Agentic AI in Autonomous Networks

Telecommunication networks are increasingly expected to operate autonomously while supporting heterogeneous services with diverse and often conflicting intents -- that is, performance objectives, constraints, and requirements specific to each service. However, transforming high-level intents -- such as ultra-low latency, high throughput, or energy efficiency -- into concrete control actions (i.e., low-level actuator commands) remains beyond the capability of existing heuristic approaches. This work introduces an Agentic AI system for intent-driven autonomous networks, structured around three specialized agents. A supervisory interpreter agent, powered by language models, performs both lexical parsing of intents into executable optimization templates and cognitive refinement based on feedback, constraint feasibility, and evolving network conditions. An optimizer agent converts these templates into tractable optimization problems, analyzes trade-offs, and derives preferences across objectives. Lastly, a preference-driven controller agent, based on multi-objective reinforcement learning, leverages these preferences to operate near the Pareto frontier of network performance that best satisfies the original intent. Collectively, these agents enable networks to autonomously interpret, reason over, adapt to, and act upon diverse intents and network conditions in a scalable manner.

FlowSteer: Interactive Agentic Workflow Orchestration via End-to-End Reinforcement Learning

In recent years, a variety of powerful agentic workflows have been applied to solve a wide range of human problems. However, existing workflow orchestration still faces key challenges, including high manual cost, reliance on specific operators/large language models (LLMs), and sparse reward signals. To address these challenges, we propose FlowSteer, an end-to-end reinforcement learning framework that takes a lightweight policy model as the agent and an executable canvas environment, automating workflow orchestration through multi-turn interaction. In this process, the policy model analyzes execution states and selects editing actions, while the canvas executes operators and returns feedback for iterative refinement. Moreover, FlowSteer provides a plug-and-play framework that supports diverse operator libraries and interchangeable LLM backends. To effectively train this interaction paradigm, we propose Canvas Workflow Relative Policy Optimization (CWRPO), which introduces diversity-constrained rewards with conditional release to stabilize learning and suppress shortcut behaviors. Experimental results on twelve datasets show that FlowSteer significantly outperforms baselines across various tasks.

Scaling In-Context Online Learning Capability of LLMs via Cross-Episode Meta-RL

Large language models (LLMs) achieve strong performance when all task-relevant information is available upfront, as in static prediction and instruction-following problems. However, many real-world decision-making tasks are inherently online: crucial information must be acquired through interaction, feedback is delayed, and effective behavior requires balancing information collection and exploitation over time. While in-context learning enables adaptation without weight updates, existing LLMs often struggle to reliably leverage in-context interaction experience in such settings. In this work, we show that this limitation can be addressed through training. We introduce ORBIT, a multi-task, multi-episode meta-reinforcement learning framework that trains LLMs to learn from interaction in context. After meta-training, a relatively small open-source model (Qwen3-14B) demonstrates substantially improved in-context online learning on entirely unseen environments, matching the performance of GPT-5.2 and outperforming standard RL fine-tuning by a large margin. Scaling experiments further reveal consistent gains with model size, suggesting significant headroom for learn-at-inference-time decision-making agents. Code reproducing the results in the paper can be found at https://github.com/XiaofengLin7/ORBIT.

Scaling Search-Augmented LLM Reasoning via Adaptive Information Control

Search-augmented reasoning agents interleave multi-step reasoning with external information retrieval, but uncontrolled retrieval often leads to redundant evidence, context saturation, and unstable learning. Existing approaches rely on outcome-based reinforcement learning (RL), which provides limited guidance for regulating information acquisition. We propose DeepControl, a framework for adaptive information control based on a formal notion of information utility, which measures the marginal value of retrieved evidence under a given reasoning state. Building on this utility, we introduce retrieval continuation and granularity control mechanisms that selectively regulate when to continue and stop retrieval, and how much information to expand. An annealed control strategy enables the agent to internalize effective information acquisition behaviors during training. Extensive experiments across seven benchmarks demonstrate that our method consistently outperforms strong baselines. In particular, our approach achieves average performance improvements of 9.4% and 8.6% on Qwen2.5-7B and Qwen2.5-3B, respectively, over strong outcome-based RL baselines, and consistently outperforms both retrieval-free and retrieval-based reasoning methods without explicit information control. These results highlight the importance of adaptive information control for scaling search-augmented reasoning agents to complex, real-world information environments.

Self-Compression of Chain-of-Thought via Multi-Agent Reinforcement Learning

The inference overhead induced by redundant reasoning undermines the interactive experience and severely bottlenecks the deployment of Large Reasoning Models. Existing reinforcement learning (RL)-based solutions tackle this problem by coupling a length penalty with outcome-based rewards. This simplistic reward weighting struggles to reconcile brevity with accuracy, as enforcing brevity may compromise critical reasoning logic. In this work, we address this limitation by proposing a multi-agent RL framework that selectively penalizes redundant chunks, while preserving essential reasoning logic. Our framework, Self-Compression via MARL (SCMA), instantiates redundancy detection and evaluation through two specialized agents: \textbf{a Segmentation Agent} for decomposing the reasoning process into logical chunks, and \textbf{a Scoring Agent} for quantifying the significance of each chunk. The Segmentation and Scoring agents collaboratively define an importance-weighted length penalty during training, incentivizing \textbf{a Reasoning Agent} to prioritize essential logic without introducing inference overhead during deployment. Empirical evaluations across model scales demonstrate that SCMA reduces response length by 11.1\% to 39.0\% while boosting accuracy by 4.33\% to 10.02\%. Furthermore, ablation studies and qualitative analysis validate that the synergistic optimization within the MARL framework fosters emergent behaviors, yielding more powerful LRMs compared to vanilla RL paradigms.

RoboStriker: Hierarchical Decision-Making for Autonomous Humanoid Boxing

Achieving human-level competitive intelligence and physical agility in humanoid robots remains a major challenge, particularly in contact-rich and highly dynamic tasks such as boxing. While Multi-Agent Reinforcement Learning (MARL) offers a principled framework for strategic interaction, its direct application to humanoid control is hindered by high-dimensional contact dynamics and the absence of strong physical motion priors. We propose RoboStriker, a hierarchical three-stage framework that enables fully autonomous humanoid boxing by decoupling high-level strategic reasoning from low-level physical execution. The framework first learns a comprehensive repertoire of boxing skills by training a single-agent motion tracker on human motion capture data. These skills are subsequently distilled into a structured latent manifold, regularized by projecting the Gaussian-parameterized distribution onto a unit hypersphere. This topological constraint effectively confines exploration to the subspace of physically plausible motions. In the final stage, we introduce Latent-Space Neural Fictitious Self-Play (LS-NFSP), where competing agents learn competitive tactics by interacting within the latent action space rather than the raw motor space, significantly stabilizing multi-agent training. Experimental results demonstrate that RoboStriker achieves superior competitive performance in simulation and exhibits sim-to-real transfer. Our website is available at RoboStriker.

TRIP-Bench: A Benchmark for Long-Horizon Interactive Agents in Real-World Scenarios

As LLM-based agents are deployed in increasingly complex real-world settings, existing benchmarks underrepresent key challenges such as enforcing global constraints, coordinating multi-tool reasoning, and adapting to evolving user behavior over long, multi-turn interactions. To bridge this gap, we introduce \textbf{TRIP-Bench}, a long-horizon benchmark grounded in realistic travel-planning scenarios. TRIP-Bench leverages real-world data, offers 18 curated tools and 40+ travel requirements, and supports automated evaluation. It includes splits of varying difficulty; the hard split emphasizes long and ambiguous interactions, style shifts, feasibility changes, and iterative version revision. Dialogues span up to 15 user turns, can involve 150+ tool calls, and may exceed 200k tokens of context. Experiments show that even advanced models achieve at most 50\% success on the easy split, with performance dropping below 10\% on hard subsets. We further propose \textbf{GTPO}, an online multi-turn reinforcement learning method with specialized reward normalization and reward differencing. Applied to Qwen2.5-32B-Instruct, GTPO improves constraint satisfaction and interaction robustness, outperforming Gemini-3-Pro in our evaluation. We expect TRIP-Bench to advance practical long-horizon interactive agents, and GTPO to provide an effective online RL recipe for robust long-horizon training.