Recent vision-language models (VLMs) show strong reasoning capabilities through training with reinforcement learning from verifiable rewards (RLVR). Despite their impressive capabilities, current VLMs focus on a limited range of reasoning tasks, such as mathematical and logical reasoning, due to the lack of readily available verifiable reward data in broader domains. As a result, these models struggle to generalize their reasoning abilities to the wide variety of challenges encountered in real-world environments. To address this limitation, we collect and assemble a comprehensive RL-ready visual reasoning training dataset encompassing 46 datasets across 13 dimensions of 5 domains, covering a wide range of realistic scenarios such as infographic reasoning, mathematical reasoning, spatial reasoning, and general science reasoning. Based on this dataset, we propose an influence function-based data filtering strategy and a multi-round data curriculum method to iteratively strengthen general visual reasoning abilities. Using this approach, we train a general reasoning VLM, namely Vision-G1. Our 7B model achieves state-of-the-art performance across nine visual reasoning benchmarks, surpassing previous similar-sized VLMs and even GPT-4o and Gemini-1.5 Flash.

Yanbin Yin, Kun Zhou, Zhen Wang, Xiangdong Zhang, Yifei Shao, Shibo Hao, Yi Gu, Jieyuan Liu, Somanshu Singla, Tianyang Liu, Eric P. Xing, Zhengzhong Liu, Haojian Jin, Zhiting Hu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

The ability to generate diverse solutions to a given problem is a hallmark of human creativity. This divergent reasoning is also crucial for machines, enhancing their robustness and enabling them to assist humans in many applications such as scientific discovery. However, existing approaches to multi-step reasoning with large language models (LLMs) have mostly focused only on reasoning accuracy, without further discovering more diverse valid solutions. For example, supervised fine-tuning improves reasoning quality but requires vast labeled data, while reward-maximizing reinforcement learning finds top-reward solutions while neglecting the solution diversity. To fill this gap, we propose Flow of Reasoning (FoR), an efficient diversity-seeking LLM finetuning method aimed at improving reasoning quality and diversity with minimal data. FoR formulates multi-step LLM reasoning as a Markovian flow on a DAG-structured reasoning graph. This formulation allows us to incorporate and adapt principled GFlowNet approaches, for finetuning LLMs to sample divergent paths with probabilities proportional to the (unnormalized) reward of target problems. Extensive experiments show that, with limited training examples (e.g., 15 examples), FoR enables the discovery of diverse, creative, high-quality solutions, greatly outperforming a wide range of existing inference and training methods across six challenging reasoning tasks, including BlocksWorld (embodied reasoning), Game24 (math puzzle solving), Rubik's Cube (spatial reasoning), 1D-ARC (abstraction reasoning), GSM8k (math reasoning), and ProntoQA (logical reasoning). Code is available at https://github.com/Yu-Fangxu/FoR.

Reinforcement learning (RL) has shown promise in enhancing large language model (LLM) reasoning, yet progress towards broader capabilities is limited by the availability of high-quality, multi-domain datasets. This work introduces \ours, a 92K RL-for-reasoning dataset designed to address this gap, covering six reasoning domains: Math, Code, Science, Logic, Simulation, and Tabular, each with corresponding verifiers. We build \ours via a careful data-curation pipeline, including sourcing, deduplication, reward design, and domain-specific and difficulty-based filtering, to facilitate the systematic investigation of cross-domain RL generalization. Our study using \ours suggests the efficacy of a simple mixed-domain RL training approach and reveals several key aspects affecting cross-domain transferability. We further train two models {\ours}-7B and {\ours}-32B purely with RL on our curated data and observe largely improved performance over leading open RL reasoning model baselines, with gains of 7.3\% and 7.8\% respectively on an extensive 17-task, six-domain evaluation suite. We are releasing our dataset, code, and evaluation suite to the community, aiming to support further research and development of more general RL-enhanced reasoning models.

Large Language Models (LLMs) have demonstrated remarkable performance in many applications, including challenging reasoning problems via chain-of-thought (CoT) techniques that generate ``thinking tokens'' before answering the questions. While existing theoretical works demonstrate that CoT with discrete tokens boosts the capability of LLMs, recent work on continuous CoT lacks a theoretical understanding of why it outperforms discrete counterparts in various reasoning tasks, such as directed graph reachability, a fundamental graph reasoning problem that includes many practical domain applications as special cases. In this paper, we prove that a two-layer transformer with $D$ steps of continuous CoT can solve the directed graph reachability problem, where $D$ is the diameter of the graph, while the best known result of constant-depth transformers with discrete CoT requires $O(n^2)$ decoding steps where $n$ is the number of vertices ($D<n$). In our construction, each continuous thought vector is a superposition state that encodes multiple search frontiers simultaneously (i.e., parallel breadth-first search (BFS)), while discrete CoT must choose a single path sampled from the superposition state, which leads to a sequential search that requires many more steps and may be trapped in local solutions. We also performed extensive experiments to verify that our theoretical construction aligns well with the empirical solution obtained via training dynamics. Notably, encoding of multiple search frontiers as a superposition state automatically emerges in training continuous CoT, without explicit supervision to guide the model to explore multiple paths simultaneously.

Benefiting from Wireless Power Transfer (WPT) technology, Wireless Rechargeable Sensor Networks (WRSNs) effectively address the lifetime bottleneck of sensor nodes, enabling them to work perpetually. Most state-of-the-art studies assume that all WRSNs’ information is known or precise in advance. However, sensor nodes may be deployed randomly in a large-scale area, and some critical information (such as node location) may be unavailable or difficult to obtain precisely. In this work, we eliminate the effect of uncertain or imprecise node location and formalize theMaximizingChargingEnergy utility for uncertain location nodesproblem (i.e., MCE problem). With magnetic resonance coupling and beamforming technologies, we propose a novel node localization method to determine precise node location information. In addition, we present a reinforcement learning framework and a charging path scheduling method to maximize charging energy. To validate the effectiveness of our proposed scheme in real-world scenarios, we conduct test-bed experiments. The results demonstrate that our approach significantly improves charging efficiency by an average of 20.9% in a large-scale network, even when the locations of sensors are entirely unknown.

Benefitting from the discovery of wireless power transfer (WPT) technology, the wireless rechargeable sensor network (WRSN) has become a promising way for lifetime extension for wireless sensor networks. In practical WRSN scenarios, obstacles can be found almost everywhere. Most state-of-the-art researches believe that obstacles will always degrade signal strength, and omit the influence of obstacles for simplifying the computation process. However, overlooking the positive impacts of obstacles on signal propagation is inconsistent with the intrinsic features of electromagnetic waves. To address this issue, in this paper, we explore the wireless signal propagation process and provide a theoretical charging model to enhance the charging efficiency by leveraging obstacles. Through utilizing the concept of the Fresnel Zone model, we re-formalize the wireless charging model and discretize the charging area and charging time to determine the best charging locations as well as charging duration. We model the chargingEfficiencyMaximization withObstacles (EMO) problem as a submodular function maximization problem and propose a cost-efficient algorithm to solve it. Finally, test-bed experiments and extensive simulations are both conducted to verify that our schemes outperform baseline algorithms by$33.46\%$on average in charging efficiency improvement.

Large language models (LLMs) have shown remarkable reasoning capabilities, particularly with chain-of-thought (CoT) prompting.However, LLMs sometimes still struggle with problems that are easy for humans, such as generating action plans to achieve given goals in an environment, or performing complex math or logical reasoning.The deficiency stems from the key fact that LLMs lack an internal world model to predict the world state (e.g., environment status, intermediate variable values) and simulate long-term outcomes of actions.This prevents LLMs from performing deliberate planning akin to human brains, which involves exploring alternative reasoning paths, anticipating future states and rewards, and iteratively refining existing reasoning steps.To overcome the limitations, we propose a new LLM reasoning framework, Reasoning via Planning (RAP).RAP repurposes the LLM as both a world model and a reasoning agent, and incorporates a principled planning algorithm based on Monte Carlo Tree Search for strategic exploration in the vast reasoning space.During reasoning, the LLM (as agent) incrementally builds a reasoning tree under the guidance of the LLM (as world model) and rewards, and efficiently obtains a high-reward reasoning path with a proper balance between exploration vs. exploitation.We apply RAP to various challenging reasoning problems including plan generation, math reasoning, and logical inference, and demonstrate its superiority over strong baselines.RAP with LLaMA-33B even surpasses CoT with GPT-4, achieving 33% relative improvement in a plan generation setting. 1

Integrating large language models (LLMs) with various tools has led to increased attention in the field. Existing approaches either involve fine-tuning the LLM, which is both computationally costly and limited to a fixed set of tools, or prompting LLMs by in-context tool demonstrations. Although the latter method offers adaptability to new tools, it struggles with the inherent context length constraint of LLMs when many new tools are presented, and mastering a new set of tools with few-shot examples remains challenging, resulting in suboptimal performance. To address these limitations, we propose a novel solution, named **ToolkenGPT**, wherein LLMs effectively learn to master tools as predicting tokens through **tool embeddings** for solving complex tasks. In this framework, each tool is transformed into vector embeddings and plugged into the language model head. Once the function is triggered during text generation, the LLM enters a special function mode to execute the tool calls. Our experiments show that function embeddings effectively help LLMs understand tool use and improve on several tasks, including numerical reasoning, knowledge-based question answering and embodied decision-making.

Wireless Rechargeable Sensor Networks (WRSNs) with perpetual network lifetime have been used in many Internet of Things (IoT) applications, like oceanic monitoring and precision agriculture. Rechargeable sensors, together with an Unmanned Aerial Vehicle (UAV), are collaboratively employed for fulfilling periodic coverage missions. However, traditional coverage solutions are normally based on static deployment of sensors and not suitable for such novel coverage requirements. In this paper, we propose the concept of Period-Area Coverage (PAC) problem, which requires the data of the overall area must be collected/monitored periodically. To solve the PAC problem, we employ a UAV that simultaneously acts as a mobile charger and sensor. It is responsible for charging nearly exhausted sensors and sensing vacant regions to realize complete event monitoring. To maximize the energy efficiency of the UAV, we propose a heuristic hexagon-based scheduling algorithm (HSA) which can also balance energy consumption. Furthermore, we develop an emergent node charging scheduling method to prevent node exhaustion, and introduce a grid-based boustrophedon scheduling algorithm (GBSA) to reduce the complexity. Finally, we present a charging re-allocation mechanism to further enhance energy efficiency. Extensive simulations demonstrate that the proposed schemes can solve the PAC problem and enhance energy efficiency by at least 18.2% compared to prior arts. Test-bed experiments conducted both in agriculture and oceanic monitoring applications validate the applicability of the proposed scheme in practical scenarios.

DocRED is a widely used dataset for documentlevel relation extraction.In the large-scale annotation, a recommend-revise scheme is adopted to reduce the workload.Within this scheme, annotators are provided with candidate relation instances from distant supervision, and they then manually supplement and remove relational facts based on the recommendations.However, when comparing Do-cRED with a subset relabeled from scratch, we find that this scheme results in a considerable amount of false negative samples and an obvious bias towards popular entities and relations.Furthermore, we observe that the models trained on DocRED have low recall on our relabeled dataset and inherit the same bias in the training data.Through the analysis of annotators' behaviors, we figure out the underlying reason for the problems above: the scheme actually discourages annotators from supplementing adequate instances in the revision phase.We appeal to future research to take into consideration the issues with the recommend-revise scheme when designing new models and annotation schemes.The relabeled dataset is released at https://github. com/AndrewZhe/Revisit-DocRED, to serve as a more reliable test set of document RE models.

Reasoning over commonsense knowledge bases (CSKBs) whose elements are in the form of free-text is an important yet hard task in NLP.While CSKB completion only fills the missing links within the domain of the CSKB, CSKB population is alternatively proposed with the goal of reasoning unseen assertions from external resources.In this task, CSKBs are grounded to a large-scale eventuality (activity, state, and event) graph to discriminate whether novel triples from the eventuality graph are plausible or not.However, existing evaluations on the population task are either not accurate (automatic evaluation with randomly sampled negative examples) or of small scale (human annotation).In this paper, we benchmark the CSKB population task with a new large-scale dataset by first aligning four popular CSKBs, and then presenting a highquality human-annotated evaluation set to probe neural models' commonsense reasoning ability.We also propose a novel inductive commonsense reasoning model that reasons over graphs.Experimental results show that generalizing commonsense reasoning on unseen assertions is inherently a hard task.Models achieving high accuracy during training perform poorly on the evaluation set, with a large gap between human performance.