Reinforcement Learning with Verifiable Rewards (RLVR) has demonstrated success in enhancing LLM reasoning capabilities, but remains limited to single-turn interactions without tool integration. While recent Agentic Reinforcement Learning with Tool use (ARLT) approaches have emerged to address multi-turn tool interactions, existing works develop task-specific codebases that suffer from fragmentation, synchronous execution bottlenecks, and limited extensibility across domains. These inefficiencies hinder broader community adoption and algorithmic innovation. We introduce VerlTool, a unified and modular framework that addresses these limitations through systematic design principles. VerlTool provides four key contributions: (1) upstream alignment with VeRL ensuring compatibility and simplified maintenance, (2) unified tool management via standardized APIs supporting diverse modalities including code execution, search, SQL databases, and vision processing, (3) asynchronous rollout execution achieving near 2$\times$ speedup by eliminating synchronization bottlenecks, and (4) comprehensive evaluation demonstrating competitive performance across 6 ARLT domains. Our framework formalizes ARLT as multi-turn trajectories with multi-modal observation tokens (text/image/video), extending beyond single-turn RLVR paradigms. We train and evaluate models on mathematical reasoning, knowledge QA, SQL generation, visual reasoning, web search, and software engineering tasks, achieving results comparable to specialized systems while providing unified training infrastructure. The modular plugin architecture enables rapid tool integration requiring only lightweight Python definitions, significantly reducing development overhead and providing a scalable foundation for tool-augmented RL research. Our code is open-sourced at https://github.com/TIGER-AI-Lab/verl-tool.

Rockim Prowell, 34, faces murder charges for allegedly killing two men and attempting to murder another after meeting victims through an online dating app in California.

Fox News Flash top sports headlines are here. Check out what's clicking on Foxnews.com. The U.S. men's national team is vying for the coveted CONCACAF Gold Cup winners trophy. But, as the USMNT prepared for Wednesday's semifinal match against Guatemala, a flying object caused a disruption at the team's training grounds. An unidentified party was believed to have been operating what appeared to be a drone in the vicinity of the team's training facility in St. Louis, CBS Sports reported.

Fox News Flash top sports headlines are here. Check out what's clicking on Foxnews.com. As the 2024 Summer Olympics in Paris draws closer, a high-profile announcer is set to lend his voice to the Games coverage. But longtime NFL play-by-play broadcaster Al Michaels will not be doing the heavy lifting. An artificial intelligence generated version of Michaels' voice will be used for Olympic recaps, NBC announced.

Information-seeking questions in long-form question answering (LFQA) often prove misleading due to ambiguity or false presupposition in the question. While many existing approaches handle misleading questions, they are tailored to limited questions, which are insufficient in a real-world setting with unpredictable input characteristics. In this work, we propose PreWoMe, a unified approach capable of handling any type of information-seeking question. The key idea of PreWoMe involves extracting presuppositions in the question and exploiting them as working memory to generate feedback and action about the question. Our experiment shows that PreWoMe is effective not only in tackling misleading questions but also in handling normal ones, thereby demonstrating the effectiveness of leveraging presuppositions, feedback, and action for real-world QA settings.

Pre-trained large language models (LLMs) have recently achieved better generalization and sample efficiency in autonomous web automation. However, the performance on real-world websites has still suffered from (1) open domainness, (2) limited context length, and (3) lack of inductive bias on HTML. We introduce WebAgent, an LLM-driven agent that learns from self-experience to complete tasks on real websites following natural language instructions. WebAgent plans ahead by decomposing instructions into canonical sub-instructions, summarizes long HTML documents into task-relevant snippets, and acts on websites via Python programs generated from those. We design WebAgent with Flan-U-PaLM, for grounded code generation, and HTML-T5, new pre-trained LLMs for long HTML documents using local and global attention mechanisms and a mixture of long-span denoising objectives, for planning and summarization. We empirically demonstrate that our modular recipe improves the success on real websites by over 50%, and that HTML-T5 is the best model to solve various HTML understanding tasks; achieving 18.7% higher success rate than the prior method on MiniWoB web automation benchmark, and SoTA performance on Mind2Web, an offline task planning evaluation.

Knowledge graphs comprise structural and textual information to represent knowledge. To predict new structural knowledge, current approaches learn representations using both types of information through knowledge graph embeddings and language models. These approaches commit to a single pre-trained language model. We hypothesize that heterogeneous language models may provide complementary information not exploited by current approaches. To investigate this hypothesis, we propose a unified framework that integrates multiple representations of structural knowledge and textual information. Our approach leverages hypercomplex algebra to model the interactions between (i) graph structural information and (ii) multiple text representations. Specifically, we utilize Dihedron models with 4*D dimensional hypercomplex numbers to integrate four different representations: structural knowledge graph embeddings, word-level representations (e.g., Word2vec and Fast-Text), sentence-level representations (using a sentence transformer), and document-level representations (using FastText or Doc2vec). Our unified framework score the plausibility of labeled edges via Dihedron products, thus modeling pairwise interactions between the four representations. Extensive experimental evaluations on standard benchmark datasets confirm our hypothesis showing the superiority of our two new frameworks for link prediction tasks.

Language models retain a significant amount of world knowledge from their pre-training stage. This allows knowledgeable models to be applied to knowledge-intensive tasks prevalent in information retrieval, such as ranking or question answering. Understanding how and which factual information is acquired by our models is necessary to build responsible models. However, limited work has been done to understand the effect of pre-training tasks on the amount of knowledge captured and forgotten by language models during pre-training. Building a better understanding of knowledge acquisition is the goal of this paper. Therefore, we utilize a selection of pre-training tasks to infuse knowledge into our model. In the following steps, we test the model's knowledge retention by measuring its ability to answer factual questions. Our experiments show that masking entities and principled masking of correlated spans based on pointwise mutual information lead to more factual knowledge being retained than masking random tokens. Our findings demonstrate that, like the ability to perform a task, the (factual) knowledge acquired from being trained on that task is forgotten when a model is trained to perform another task (catastrophic forgetting) and how to prevent this phenomenon. To foster reproducibility, the code, as well as the data used in this paper, are openly available.

Fox News correspondent Grady Trimble has the latest on fears the technology will spiral out of control on'Special Report.' Introductions of superstars are among the main attractions of a WWE event. The entrances are a big part of the superstars' characters and have become something fans have come to rely on seeing. But with AI tools such as ChatGPT becoming more widely used, more wrestlers could opt to allow technology to handle their entrances. In this photo provided by WWE, Roman Reigns, center, holds up his WWE heavyweight and universal championship belts after defeating Cody Rhodes in the main event of WrestleMania 39 April 2, 2023, at SoFi Stadium in Inglewood, Calif.

Closed-book question answering (QA) requires a model to directly answer an open-domain question without access to any external knowledge. Prior work on closed-book QA either directly finetunes or prompts a pretrained language model (LM) to leverage the stored knowledge. However, they do not fully exploit the parameterized knowledge. To address this issue, we propose a two-stage, closed-book QA framework which employs a coarse-to-fine approach to extract relevant knowledge and answer a question. Our approach first generates a related context for a given question by prompting a pretrained LM. We then prompt the same LM for answer prediction using the generated context and the question. Additionally, to eliminate failure caused by context uncertainty, we marginalize over generated contexts. Experimental results on three QA benchmarks show that our method significantly outperforms previous closed-book QA methods (e.g. exact matching 68.6% vs. 55.3%), and is on par with open-book methods that exploit external knowledge sources (e.g. 68.6% vs. 68.0%). Our method is able to better exploit the stored knowledge in pretrained LMs without adding extra learnable parameters or needing finetuning, and paves the way for hybrid models that integrate pretrained LMs with external knowledge.