Yawei Wang

Linbo Liu, Guande Wu, Han Ding et al.

Large language model agents rely on effective model context to obtain task-relevant information for decision-making. Many existing context engineering approaches primarily rely on the context generated from the past experience and retrieval mechanisms that reuse these context. However, retrieved context from past tasks must be adapted by the execution agent to fit new situations, placing additional reasoning burden on the underlying LLM. To address this limitation, we propose a generative context augmentation framework using Contrastive Learning of Experience via Agentic Reflection (CLEAR). CLEAR first employs a reflection agent to perform contrastive analysis over past execution trajectories and summarize useful context for each observed task. These summaries are then used as supervised fine-tuning data to train a context augmentation model (CAM). Then we further optimize CAM using reinforcement learning, where the reward signal is obtained by running the task execution agent. By learning to generate task-specific knowledge rather than retrieve knowledge from the past, CAM produces context that is better tailored to the current task. We conduct comprehensive evaluations on the AppWorld and WebShop benchmarks. Experimental results show that CLEAR consistently outperforms strong baselines. It improves task completion rate from 72.62% to 81.15% on AppWorld test set and averaged reward from 0.68 to 0.74 on a subset of WebShop, compared with baseline agent. Our code is publicly available at https://github.com/awslabs/CLEAR.

Wenqian Ye, Bo Yuan, Zhichao Xu et al.

We study a class of emergent misalignment in multi-agent systems (MAS), with a focus on automated workflows, which we refer to agentic misalignment. Although these systems can solve complex tasks, they often fail because agents act according to implicit proxy utilities that do not align with the intended human goals. We formally define these behaviors and analyze them within a Bayesian framework, showing that generic utilities naturally lead to posterior collapse of agents in automated workflows. To address this issue, we propose Agentic Evidence Attribution (AEA), a novel alignment paradigm that improves agent posteriors using context-specific evidence. AEA reasons over agent actions and provides structured evidence to correct misaligned behavior during collaboration. To better understand the role of evidence, we study two instantiations of AEA: self-reflection (internal evidence from the model) and weak-to-strong generalization (external evidence on the agentic trajectory). We show that a small evidence model effectively aligns the MAS by providing orthogonal failure attribution. Our results clarify the sources of agentic misalignment in automated workflows and show that evidence-based alignment can effectively improve agent collaboration and leads to reliable multi-agent systems built on automated workflows.

Jiongxiao Wang, Qiaojing Yan, Yawei Wang et al.

Large Language Model (LLM)-based agents have demonstrated remarkable capabilities in complex reasoning and multi-turn interactions but struggle to continuously improve and adapt when deployed in new environments. One promising approach is implementing skill libraries that allow agents to learn, validate, and apply new skills. However, current skill library approaches rely primarily on LLM prompting, making consistent skill library implementation challenging. To overcome these challenges, we propose a Reinforcement Learning (RL)-based approach to enhance agents' self-improvement capabilities with a skill library. Specifically, we introduce Skill Augmented GRPO for self-Evolution (SAGE), a novel RL framework that systematically incorporates skills into learning. The framework's key component, Sequential Rollout, iteratively deploys agents across a chain of similar tasks for each rollout. As agents navigate through the task chain, skills generated from previous tasks accumulate in the library and become available for subsequent tasks. Additionally, the framework enhances skill generation and utilization through a Skill-integrated Reward that complements the original outcome-based rewards. Experimental results on AppWorld demonstrate that SAGE, when applied to supervised-finetuned model with expert experience, achieves 8.9% higher Scenario Goal Completion while requiring 26% fewer interaction steps and generating 59% fewer tokens, substantially outperforming existing approaches in both accuracy and efficiency.

Dongkyu Derek Cho, Huan Song, Arijit Ghosh Chowdhury et al.

Fine-tuning large language models (LLMs) for downstream tasks typically exhibit a fundamental safety-capability tradeoff, where improving task performance degrades safety alignment even on benign datasets. This degradation persists across standard approaches including supervised finetuning (SFT) and reinforcement learning from human feedback (RLHF). While reinforcement learning with verifiable rewards (RLVR) has emerged as a promising alternative that optimizes models on objectively measurable tasks, its safety implications remain unexplored. We present the first comprehensive theoretical and empirical analysis of safety properties in RLVR. Theoretically, we derive upper bounds on safety drift under KL-constrained optimization and prove conditions under which safety degradation is eliminated. Empirically, we conduct extensive experiments across five adversarial safety benchmarks, demonstrating that RLVR can simultaneously enhance reasoning capabilities while maintaining or improving safety guardrails. Our comprehensive ablation studies examine the effects of optimization algorithms, model scale, and task domains. Our findings challenge the prevailing assumption of an inevitable safety capability trade-off, and establish that a specific training methodology can achieve both objectives simultaneously, providing insights for the safe deployment of reasoning-capable LLMs.

Zhichao Xu, Minheng Wang, Yawei Wang et al.

Retrieval-augmented generation (RAG) systems trained using reinforcement learning (RL) with reasoning are hampered by inefficient context management, where long, noisy retrieved documents increase costs and degrade performance. We introduce RECON (REasoning with CONdensation), a framework that integrates an explicit summarization module to compress evidence within the reasoning loop. Our summarizer is trained via a two-stage process: relevance pretraining on QA datasets, followed by multi-aspect distillation from proprietary LLMs to ensure factuality and clarity. Integrated into the Search-R1 pipeline, RECON reduces total context length by 35\%, leading to improved training speed and inference latency, while simultaneously improving RAG performance on downstream QA benchmarks. Notably, it boosts the average EM score of the 3B model by 14.5\% and the 7B model by 3.0\%, showing particular strength in multi-hop QA. RECON demonstrates that learned context compression is essential for building practical, scalable, and performant RAG systems. Our code implementation is made available at https://github.com/allfornancy/RECON.

Kiran Ramnath, Kang Zhou, Sheng Guan et al.

Since the advent of large language models (LLMs), prompt engineering has been a crucial step for eliciting desired responses for various Natural Language Processing (NLP) tasks. However, prompt engineering remains an impediment for end users due to rapid advances in models, tasks, and associated best practices. To mitigate this, Automatic Prompt Optimization (APO) techniques have recently emerged that use various automated techniques to help improve the performance of LLMs on various tasks. In this paper, we present a comprehensive survey summarizing the current progress and remaining challenges in this field. We provide a formal definition of APO, a 5-part unifying framework, and then proceed to rigorously categorize all relevant works based on their salient features therein. We hope to spur further research guided by our framework.

Guande Wu, Huan Song, Yawei Wang et al.

Reasoning is increasingly crucial for various tasks. While chain-of-thought prompting enables large language models to leverage reasoning effectively, harnessing the reasoning capabilities of Vision-Language Models (VLMs) remains challenging. To solve this problem, we propose a novel self-distillation framework that enhances the reasoning capabilities of the model. The proposed framework introduces several key innovations. We start by employing a prompt library tailored to visual reasoning tasks to generate diverse in-context questions and utilize a two-step reasoning procedure to derive reasoning-guided responses. These responses are then used for self-distillation, enabling the model to internalize the reasoning process. Additionally, we improve the model architecture with several innovative components, including an intervention adapter for efficient parameter updates, a cross-modal skip connection to facilitate information exchange between modalities, and an ensemble learning algorithm to integrate diverse reasoning from multiple in-context questions. Extensive experiments show that our method significantly improves the baseline performance across five VQA datasets.

Jiazheng Li, Yawei Wang, David Yan et al.

Large Language Models (LLMs) have demonstrated remarkable capabilities, enabling language agents to excel at single-turn tasks. However, their application to complex, multi-step, and long-horizon tasks remains challenging. While reinforcement learning (RL) offers a promising avenue for addressing these challenges, mainstream approaches typically rely solely on sparse, outcome-based rewards, a limitation that becomes especially problematic for group-based RL algorithms lacking critic models, such as Group Relative Policy Optimization (GRPO). In such methods, uniformly rewarding or penalizing all actions within a trajectory can lead to training instability and suboptimal policies, because beneficial and detrimental actions are often entangled across multi-step interactions. To address this challenge, we propose SALT, a novel and lightweight framework that provides a finer-grained advantage assignment, derived solely from outcome rewards. We achieve this by constructing a graph from trajectories of the same prompt, which allows us to quantify the quality of each step and assign advantages accordingly. Crucially, SALT is designed as a plug-and-play module that seamlessly integrates with existing group-based RL algorithms, requiring no modifications to the rollout procedure and introducing negligible computational overhead. Extensive experiments on the WebShop, ALFWorld, and AppWorld benchmarks with various model sizes demonstrate that SALT consistently improves performance. We also conduct a thorough analysis to validate the design choices behind SALT and offer actionable insights.

Yijun Tian, Shaoyu Chen, Zhichao Xu et al.

The development of state-of-the-art large language models is commonly understood as a two-stage process involving pre-training and post-training. We point out the need for an additional intermediate stage called reinforcement mid-training with potential for strong performance gains. In this paper, we formally define the problem and identify three key challenges: (1) inefficient training due to excessive reasoning steps, (2) disregard of the imbalanced token entropy distribution, and (3) underutilization of token information. To address these challenges, we propose RMT, a framework for efficient, adaptive, and unified reinforcement mid-training with various innovative components. In particular, we first introduce a dynamic token budget mechanism that constrains unnecessary reasoning steps and mitigates model overthinking. Next, we design a curriculum-based adaptive sampling method that fosters a progressive learning trajectory from easy to hard tokens. Finally, we present a dual training strategy that combines reinforcement learning with next-token prediction, ensuring targeted learning on key tokens and full exploitation of all token information. Extensive experiments demonstrate the superiority of RMT over state-of-the-art methods, achieving up to +64.91% performance improvement with only 21% of the reasoning length in language modeling. We also show that checkpoints obtained after reinforcement mid-training can benefit the subsequent post-training, yielding up to +18.76% improvement in the mathematical domain.

Kazi Tasnim Zinat, Yun Zhou, Xiang Lyu et al.

Inferring causal relationships between event pairs in a temporal sequence is applicable in many domains such as healthcare, manufacturing, and transportation. Most existing work on causal inference primarily focuses on event types within the designated domain, without considering the impact of exogenous out-of-domain interventions. In real-world settings, these out-of-domain interventions can significantly alter causal dynamics. To address this gap, we propose a new causal framework to define average treatment effect (ATE), beyond independent and identically distributed (i.i.d.) data in classic Rubin's causal framework, to capture the causal relation shift between events of temporal process under out-of-domain intervention. We design an unbiased ATE estimator, and devise a Transformer-based neural network model to handle both long-range temporal dependencies and local patterns while integrating out-of-domain intervention information into process modeling. Extensive experiments on both simulated and real-world datasets demonstrate that our method outperforms baselines in ATE estimation and goodness-of-fit under out-of-domain-augmented point processes.

Zhilin Wang, Qin Hu, Minghui Xu et al.

With the technological advances in machine learning, effective ways are available to process the huge amount of data generated in real life. However, issues of privacy and scalability will constrain the development of machine learning. Federated learning (FL) can prevent privacy leakage by assigning training tasks to multiple clients, thus separating the central server from the local devices. However, FL still suffers from shortcomings such as single-point-failure and malicious data. The emergence of blockchain provides a secure and efficient solution for the deployment of FL. In this paper, we conduct a comprehensive survey of the literature on blockchained FL (BCFL). First, we investigate how blockchain can be applied to federal learning from the perspective of system composition. Then, we analyze the concrete functions of BCFL from the perspective of mechanism design and illustrate what problems blockchain addresses specifically for FL. We also survey the applications of BCFL in reality. Finally, we discuss some challenges and future research directions.

Wenwen Gonga, Yulan Zhang, Xuyun Zhang et al.

With the ever-increasing prevalence of web APIs (Application Programming Interfaces) in enabling smart software developments, finding and composing a list of existing web APIs that can corporately fulfil the software developers' functional needs have become a promising way to develop a successful mobile app, economically and conveniently. However, the big volume and diversity of candidate web APIs put additional burden on the app developers' web APIs selection decision-makings, since it is often a challenging task to simultaneously guarantee the diversity and compatibility of the finally selected a set of web APIs. Considering this challenge, a Diversity-aware and Compatibility-driven web APIs Recommendation approach, namely DivCAR, is put forward in this paper. First, to achieve diversity, DivCAR employs random walk sampling technique on a pre-built correlation graph to generate diverse correlation subgraphs. Afterwards, with the diverse correlation subgraphs, we model the compatible web APIs recommendation problem to be a minimum group Steiner tree search problem. Through solving the minimum group Steiner tree search problem, manifold sets of compatible and diverse web APIs ranked are returned to the app developers. At last, we design and enact a set of experiments on a real-world dataset crawled from www.programmableWeb.com. Experimental results validate the effectiveness and efficiency of our proposed DivCAR approach in balancing the web APIs recommendation diversity and compatibility.

Yawei Wang, Xiu Li

For adversarial imitation learning algorithms (AILs), no true rewards are obtained from the environment for learning the strategy. However, the pseudo rewards based on the output of the discriminator are still required. Given the implicit reward bias problem in AILs, we design several representative reward function shapes and compare their performances by large-scale experiments. To ensure our results' reliability, we conduct the experiments on a series of Mujoco and Box2D continuous control tasks based on four different AILs. Besides, we also compare the performance of various reward function shapes using varying numbers of expert trajectories. The empirical results reveal that the positive logarithmic reward function works well in typical continuous control tasks. In contrast, the so-called unbiased reward function is limited to specific kinds of tasks. Furthermore, several designed reward functions perform excellently in these environments as well.

Reza Bayani, Saeed D. Manshadi, Guangyi Liu et al.

A total 19% of generation capacity in California is offered by PV units and over some months, more than 10% of this energy is curtailed. In this research, a novel approach to reduce renewable generation curtailments and increasing system flexibility by means of electric vehicles' charging coordination is represented. The presented problem is a sequential decision making process, and is solved by fitted Q-iteration algorithm which unlike other reinforcement learning methods, needs fewer episodes of learning. Three case studies are presented to validate the effectiveness of the proposed approach. These cases include aggregator load following, ramp service and utilization of non-deterministic PV generation. The results suggest that through this framework, EVs successfully learn how to adjust their charging schedule in stochastic scenarios where their trip times, as well as solar power generation are unknown beforehand.

Ming Zhang, Yawei Wang, Xiaoteng Ma et al.

The generative adversarial imitation learning (GAIL) has provided an adversarial learning framework for imitating expert policy from demonstrations in high-dimensional continuous tasks. However, almost all GAIL and its extensions only design a kind of reward function of logarithmic form in the adversarial training strategy with the Jensen-Shannon (JS) divergence for all complex environments. The fixed logarithmic type of reward function may be difficult to solve all complex tasks, and the vanishing gradients problem caused by the JS divergence will harm the adversarial learning process. In this paper, we propose a new algorithm named Wasserstein Distance guided Adversarial Imitation Learning (WDAIL) for promoting the performance of imitation learning (IL). There are three improvements in our method: (a) introducing the Wasserstein distance to obtain more appropriate measure in the adversarial training process, (b) using proximal policy optimization (PPO) in the reinforcement learning stage which is much simpler to implement and makes the algorithm more efficient, and (c) exploring different reward function shapes to suit different tasks for improving the performance. The experiment results show that the learning procedure remains remarkably stable, and achieves significant performance in the complex continuous control tasks of MuJoCo.