Post-Training Multimodal Large Language Models (MLLMs) to build interactive agents holds promise across domains such as computer-use, web navigation, and robotics. A key challenge in scaling such post-training is lack of high-quality downstream agentic task datasets with tasks that are diverse, feasible, and verifiable. Existing approaches for task generation rely heavily on human annotation or prompting MLLM with limited downstream environment information, which is either costly or poorly scalable as it yield tasks with limited coverage. To remedy this, we present AutoPlay, a scalable pipeline for task generation that explicitly explores interactive environments to discover possible interactions and current state information to synthesize environment-grounded tasks. AutoPlay operates in two stages: (i) an exploration phase, where an MLLM explorer agent systematically uncovers novel environment states and functionalities, and (ii) a task generation phase, where a task generator leverages exploration trajectories and a set of task guideline prompts as context to synthesize diverse, executable, and verifiable tasks. We show AutoPlay generates 20k tasks across 20 Android applications and 10k tasks across 13 applications Ubuntu applications to train mobile-use and computer-use agents. AutoPlay generated tasks enable large-scale task demonstration synthesis without human annotation by employing an MLLM task executor and verifier. This data enables training MLLM-based UI agents that improve success rates up to $20.0\%$ on mobile-use and $10.9\%$ on computer-use scenarios. In addition, AutoPlay generated tasks combined with MLLM verifier-based rewards enable scaling reinforcement learning training of UI agents, leading to an additional $5.7\%$ gain. coverage. These results establish AutoPlay as a scalable approach for post-training capable MLLM agents reducing reliance on human annotation.

Large Language Models (LLMs) are starting to be profiled as one of the most significant disruptions in the Software Testing field. Specifically, they have been successfully applied in software testing tasks such as generating test code, or summarizing documentation. This potential has attracted hundreds of researchers, resulting in dozens of new contributions every month, hardening researchers to stay at the forefront of the wave. Still, to the best of our knowledge, no prior work has provided a structured vision of the progress and most relevant research trends in LLM-based testing. In this article, we aim to provide a roadmap that illustrates its current state, grouping the contributions into different categories, and also sketching the most promising and active research directions for the field. To achieve this objective, we have conducted a semi-systematic literature review, collecting articles and mapping them into the most prominent categories, reviewing the current and ongoing status, and analyzing the open challenges of LLM-based software testing. Lastly, we have outlined several expected long-term impacts of LLMs over the whole software testing field.

The current paradigm for reasoning in large language models (LLMs) involves models "thinking out loud" via a sequence of tokens, known as chain-of-thought (CoT). This approach, while effective, has several significant drawbacks. Firstly, inference requires autoregressive generation of often thousands of CoT tokens, which is slow and computationally expensive. Secondly, it constrains reasoning to the discrete space of tokens, creating an information bottleneck across reasoning steps. Thirdly, it fundamentally entangles reasoning with token generation, forcing LLMs to "think while speaking," which causes potentially short-sighted reasoning. In light of these limitations, we re-imagine reasoning in LLMs and present a new paradigm: MARCOS. In our approach, rather than autoregressively generating tokens, we model reasoning as a hidden Markov chain of continuous, high-dimensional "thoughts". Each reasoning step involves a transition of the internal thoughts, where explicit reasoning steps (which may consist of hundreds of tokens) serve as observable variables, which are windows to peek into the implicit thoughts. Since this latent process is incompatible with the standard supervised learning, we further propose a two-phase variational training scheme. Our experiments on three benchmarks demonstrate that MARCOS outperforms existing continuous reasoning methods and, for the first time, achieves performance comparable to token-based CoT, even surpassing it by 4.7% on GSM8K with up to 15.7x speedup in inference. Beyond this, MARCOS offers additional advantages, such as step-level instead of token-level control over randomness, opening significant opportunities for reinforcement learning and reasoning in LLMs.

Recently, online Reinforcement Learning with Verifiable Rewards (RLVR) has become a key paradigm for enhancing the reasoning capabilities of Large Language Models (LLMs). However, existing methods typically treat all training samples uniformly, overlooking the vast differences in problem difficulty relative to the model's current capabilities. This uniform training strategy leads to inefficient exploration of problems the model has already mastered, while concurrently lacking effective guidance on problems that are challenging its abilities the most, limiting both learning efficiency and upper-bound performance. To address this, we propose CLPO (Curriculum-guided Learning for Policy Optimization), a novel algorithm that creates a dynamic pedagogical feedback loop within the policy optimization process. The core of CLPO leverages the model's own rollout performance to conduct real-time difficulty assessment, thereby constructing an Online Curriculum. This curriculum then guides an Adaptive Problem Restructuring mechanism, where the model acts as its own teacher: it diversifies medium-difficulty problems to promote generalization and simplifies challenging problems to make them more attainable. Our approach transforms the static training procedure into a dynamic process that co-evolves with the model's capabilities. Experiments show that CLPO achieves state-of-the-art performance across eight challenging mathematical and general reasoning benchmarks, with an average pass@1 improvement of 6.96% over other methods, demonstrating its potential for more efficiently training more capable reasoning models.

Generating accurate and calibrated confidence estimates is critical for deploying LLMs in high-stakes or user-facing applications, and remains an open challenge. Prior research has often framed confidence as a problem of eliciting a model's "self-knowledge", i.e., the ability of an LLM to judge whether its own answers are correct; this approach implicitly assumes that there is some privileged information about the answer's correctness that is accessible to the model itself. However, our experiments reveal that an LLM attempting to predict the correctness of its own outputs generally performs no better than an unrelated LLM. Moreover, we hypothesize that a key factor in building a "Correctness Model" (CM) is exposure to a target model's historical predictions. We propose multiple methods to inject this historical correctness information, creating a Generalized Correctness Model (GCM). We first show that GCMs can be trained on the correctness data from many LLMs and learn patterns for correctness prediction applicable across datasets and models. We then use CMs as a lens for studying the source of correctness prediction ability and its generalization, systematically controlling their training data and finding that answer phrasing is a strong predictor for correctness. We further explore alternative methods of injecting history without training an LLM, finding that including history as in-context examples can help improve correctness prediction, and post-hoc calibration can provide complementary reductions in calibration error. We evaluate GCMs based on Qwen3-8B across 5 model families and the MMLU and TriviaQA datasets, as well as on a downstream selective prediction task, finding that reliable LLM confidence estimation is a generalizable and model-agnostic skill learned by systematically encoding correctness history rather than a model-specific skill reliant on self-introspection.

RL with Verifiable Rewards (RLVR) has emerged as a promising paradigm for improving the reasoning abilities of large language models (LLMs). Current methods rely primarily on policy optimization frameworks like PPO and GRPO, which follow generalized policy iteration that alternates between evaluating the current policy's value and improving the policy based on evaluation. While effective, they often suffer from training instability and diversity collapse, requiring complex heuristic tricks and careful tuning. We observe that standard RLVR in math reasoning can be formalized as a specialized finite-horizon Markov Decision Process with deterministic state transitions, tree-structured dynamics, and binary terminal rewards. Though large in scale, the underlying structure is simpler than general-purpose control settings for which popular RL algorithms (e.g., PPO) were developed, suggesting that several sophisticated techniques in existing methods may be reduced or even omitted. Based on this insight, we prove a surprising result: the optimal action can be recovered from the Q-function of a fixed uniformly random policy, thereby bypassing the generalized policy iteration loop and its associated heuristics. We introduce Random Policy Valuation for Diverse Reasoning (ROVER) to translate this principle into a practical and scalable algorithm for LLM math reasoning, a minimalist yet highly effective RL method that samples actions from a softmax over these uniform-policy Q-values. ROVER preserves diversity throughout training, allowing sustained exploration of multiple valid pathways. Across multiple base models and standard math reasoning benchmarks, ROVER demonstrates superior performance in both \textbf{quality} (\textbf{+8.2} on pass@1, \textbf{+16.8} on pass@256) and \textbf{diversity} (\textbf{+17.6\%}), despite its radical simplification compared to strong, complicated existing methods.

Data scarcity drives the need for more sample-efficient large language models. In this work, we use the double descent phenomenon to holistically compare the sample efficiency of discrete diffusion and autoregressive models. We show that discrete diffusion models require larger capacity and more training epochs to escape their underparameterized regime and reach the interpolation threshold. In the strongly overparameterized regime, both models exhibit similar behavior, with neither exhibiting a pronounced second descent in test loss across a large range of model sizes. Overall, our results indicate that autoregressive models are more sample-efficient on small-scale datasets, while discrete diffusion models only become competitive when given sufficient capacity and compute.

Recommender systems (RS) are widely used in e-commerce for personalized suggestions, yet their openness makes them susceptible to shilling attacks, where adversaries inject fake behaviors to manipulate recommendations. Most existing defenses emphasize user-side behaviors while overlooking item-side features such as titles and descriptions that can expose malicious intent. To address this gap, we propose a two-stage detection framework that integrates item-side semantics via large language models (LLMs). The first stage pre-screens suspicious users using low-cost behavioral criteria, and the second stage employs LLM-based auditing to evaluate semantic consistency. Furthermore, we enhance the auditing model through reinforcement fine-tuning on a lightweight LLM with carefully designed reward functions, yielding a specialized detector called SemanticShield. Experiments on six representative attack strategies demonstrate the effectiveness of SemanticShield against shilling attacks, and further evaluation on previously unseen attack methods shows its strong generalization capability. Code is available at https://github.com/FrankenstLee/SemanticShield.

LLMs increasingly rely on inference-time reasoning algorithms such as chain-of-thought and multi-branch reasoning to improve accuracy on complex tasks. Prior work has primarily focused on reducing token usage, often at the expense of accuracy, while overlooking other latency factors. We present DUCHESS, an LLM serving system that reduces computational cost and latency without sacrificing accuracy through intra-request branch orchestration guided by predictions. Within each request, DUCHESS predicts branch correctness with a lightweight linear probing model over LLM layer activations. The orchestration policy uses these predictions to decide whether to terminate a branch early, duplicate an existing branch, or continue exploring a branch. When handling multiple requests, DUCHESS can further reduce latency by prioritizing easier reasoning tasks, when request complexity can be estimated from the prompt. Experiments on three reasoning benchmarks show that DUCHESS consistently improves the token-accuracy Pareto frontier, reducing token usage by 42-63% at matched accuracy compared to self-consistency. For request serving with vLLM, DUCHESS reduces mean, median, and tail latencies by 57-81%, 58-85%, and 52-84% with First-Come-First-Served (FCFS) scheduling across three datasets, compared to self-consistency. At higher request rates, scheduling jobs by increasing predicted difficulty reduces latency further over FCFS. Large Language Models (LLMs) are widely applied across domains, including math and science problem-solving (Lewkowycz et al., 2022), coding and program analysis (Jiang et al., 2025), logical deduction (Creswell et al., 2023), and decision-making (Y ao et al., 2023).

Scalability has driven recent advances in generative modeling, yet its principles remain underexplored for adversarial learning. We investigate the scalability of Generative Adversarial Networks (GANs) through two design choices that have proven to be effective in other types of generative models: training in a compact V ariational Autoencoder latent space and adopting purely transformer-based generators and discriminators. Training in latent space enables efficient computation while preserving perceptual fidelity, and this efficiency pairs naturally with plain transformers, whose performance scales with computational budget. Building on these choices, we analyze failure modes that emerge when naively scaling GANs. Specifically, we find issues as underutilization of early layers in the generator and optimization instability as the network scales. Accordingly, we provide simple and scale-friendly solutions as lightweight intermediate supervision and width-aware learning-rate adjustment. Our experiments show that GA T, a purely transformer-based and latent-space GANs, can be easily trained reliably across a wide range of capacities (S through XL). Moreover, GA T -XL/2 achieves state-of-the-art single-step, class-conditional generation performance (FID of 2.96) on ImageNet-256 in just 40 epochs, 6 fewer epochs than strong baselines. GA T -XL/2 exhibits strong generation capability (FID 2.96) within 40 epochs, 6 fewer than 1-NFE baselines (FID 3.43), while keeping the characteristics of GANs such as latent interpolation (bottom two rows). Recent breakthroughs in generative modeling have become a central driver of progress across core areas of computer vision.