Despite their strengths, large language models (LLMs) often fail to communicate their confidence accurately, making it difficult to assess when they might be wrong and limiting their reliability. In this work, we demonstrate that reasoning models that engage in extended chain-of-thought (CoT) reasoning exhibit superior performance not only in problem-solving but also in accurately expressing their confidence. Specifically, we benchmark six reasoning models across six datasets and find that they achieve strictly better confidence calibration than their non-reasoning counterparts in 33 out of the 36 settings. Our detailed analysis reveals that these gains in calibration stem from the slow thinking behaviors of reasoning models (e.g., exploring alternative approaches and backtracking) which enable them to adjust their confidence dynamically throughout their CoT, making it progressively more accurate. In particular, we find that reasoning models become increasingly better calibrated as their CoT unfolds, a trend not observed in non-reasoning models. Moreover, removing slow thinking behaviors from the CoT leads to a significant drop in calibration. Lastly, we show that non-reasoning models also demonstrate enhanced calibration when simply guided to slow think via in-context learning, fully isolating slow thinking as the source of the calibration gains.2

We analyse how the sampling dynamics of distributions evolve in score-based diffusion models using cross-fluctuations, a centered-moment statistic from statistical physics. Specifically, we show that starting from an unbiased isotropic normal distribution, samples undergo sharp, discrete transitions, eventually forming distinct events of a desired distribution while progressively revealing finer structure. As this process is reversible, these transitions also occur in reverse, where intermediate states progressively merge, tracing a path back to the initial distribution. We demonstrate that these transitions can be detected as discontinuities in nth-order cross-fluctuations. For variance-preserving SDEs, we derive a closed-form for these cross-fluctuations that is efficiently computable for the reverse trajectory. We find that detecting these transitions directly boosts sampling efficiency, accelerates class-conditional and rare-class generation, and improves two zero-shot tasks-image classification and style transfer-without expensive grid search or retraining. We also show that this viewpoint unifies classical coupling and mixing from finite Markov chains with continuous dynamics while extending to stochastic SDEs and non Markovian samplers.

Foundation models (FMs) have shown great promise in single-cell genomics, yet current approaches, such as scGPT, Geneformer, and scFoundation, rely on centralized training and language modeling objectives that overlook the tabular nature of single-cell data and raise significant privacy concerns. We present TABULA, a foundation model designed for single-cell transcriptomics, which integrates a novel tabular modeling objective and federated learning framework to enable privacy-preserving pretraining across decentralized datasets.

Chain-of-thought (CoT) reasoning is critical for improving the interpretability and reliability of Large Vision-Language Models (LVLMs). However, existing training algorithms such as SFT, PPO, and GRPO may not generalize well across unseen reasoning tasks and heavily rely on a biased reward model. To address this challenge, we reformulate reasoning in LVLMs as posterior inference and propose a scalable training algorithm based on amortized variational inference. By leveraging diversity-seeking reinforcement learning algorithms, we introduce a novel sparse reward function for token-level learning signals that encourage diverse, high-likelihood latent CoT, overcoming deterministic sampling limitations and avoiding reward hacking. Additionally, we implement a Bayesian inference-scaling strategy that replaces costly Best-of-N and Beam Search with a marginal likelihood to efficiently rank optimal rationales and answers. We empirically demonstrate that the proposed method enhances the state-of-the-art LVLMs on seven reasoning benchmarks, in terms of effectiveness, generalization, and interpretability.

Recent advances have demonstrated the effectiveness of Reinforcement Learning (RL) in improving the reasoning capabilities of Large Language Models (LLMs). However, existing works inevitably rely on high-quality instructions and verifiable rewards for effective training, both of which are often difficult to obtain in specialized domains. In this paper, we propose Self-play Reinforcement Learning (SeRL) to bootstrap LLM training with limited initial data. Specifically, SeRL comprises two complementary modules: self-instruction and self-rewarding. The former module generates additional instructions based on the available data at each training step, employing robust online filtering strategies to ensure instruction quality, diversity, and difficulty. The latter module introduces a simple yet effective majority-voting mechanism to estimate response rewards for additional instructions, eliminating the need for external annotations. Finally, SeRL performs conventional RL based on the generated data, facilitating iterative self-play learning. Extensive experiments on various reasoning benchmarks and across different LLM backbones demonstrate that the proposed SeRL yields results superior to its counterparts and achieves performance on par with those obtained by high-quality data with verifiable rewards.

The era of foundation models has revolutionized AI research, yet Graph Foundation Models (GFMs) remain constrained by the scarcity of large-scale graph corpora. Traditional graph data synthesis techniques primarily focus on simplistic structural operations, lacking the capacity to generate semantically rich nodes with meaningful textual attributes--a critical limitation for real-world applications. While large language models (LLMs) demonstrate exceptional text generation capabilities, their direct application to graph synthesis is impeded by context window limitations, hallucination phenomena, and structural consistency challenges. To address these issues, we introduce GraphMaster--the first multi-agent framework specifically designed for graph data synthesis in data-limited environments.

This paper investigates the computational complexity of reinforcement learning within a novel linear function approximation regime, termed partial qπ-realizability. In this framework, the objective is to learn an ϵ-optimal policy with respect to a predefined policy set Π, under the assumption that all value functions corresponding to policies in Π are linearly realizable. This framework adopts assumptions that are weaker than those in the qπ-realizability setting yet stronger than those in the q -realizability setup. As a result, it provides a more practical model for reinforcement learning scenarios where function approximation naturally arise. We prove that learning an ϵ-optimal policy in this newly defined setting is computationally hard. More specifically, we establish NP-hardness under a parameterized greedy policy set (i.e., argmax) and, further, show that--unless NP = RP--an exponential lower bound (exponential in feature vector dimension) holds when the policy set contains softmax policies, under the Randomized Exponential Time Hypothesis. Our hardness results mirror those obtained in the q -realizability settings, and suggest that computational difficulty persists even when the policy class Πis expanded beyond the optimal policy, reinforcing the unbreakable nature of the computational hardness result regarding partial qπ-realizability under two important policy sets. To establish our negative result, our primary technical contribution is a reduction from two complexity problems, δ-MAX-3SAT and δ-MAX-3SAT(b), to instances of our problem settings: GLINEAR-κ-RL (under the greedy policy set) and SLINEAR-κ-RL (under the softmax policy set), respectively. Our findings indicate that positive computational results are generally unattainable in the context of partial qπ-realizability, in sharp contrast to the qπ-realizability setting under a generative access model.

We introduce Skeleton-Cache, the first training-free test-time adaptation framework for skeleton-based zero-shot action recognition (SZAR), aimed at improving model generalization to unseen actions during inference. Skeleton-Cache reformulates inference as a lightweight retrieval process over a non-parametric cache that stores structured skeleton representations, combining both global and fine-grained local descriptors. To guide the fusion of descriptor-wise predictions, we leverage the semantic reasoning capabilities of large language models (LLMs) to assign classspecific importance weights. By integrating these structured descriptors with LLMguided semantic priors, Skeleton-Cache dynamically adapts to unseen actions without any additional training or access to training data. Extensive experiments on NTURGB+D 60/120 and PKU-MMDII demonstrate that Skeleton-Cache consistently boosts the performance of various SZAR backbones under both zeroshot and generalized zero-shot settings.

General intelligence requires systems that acquire new skills efficiently and generalize beyond their training distributions. Although program synthesis approaches have strong generalization power, they face scaling issues due to the large combinatorial spaces that quickly render them impractical, requiring human-generated DSLs or pre-trained priors to narrow this search space. On the other hand, deep learning methods have had high successes, but they lack structured test-time adaptation and rely on heavy stochastic sampling or expensive gradient updates for fine-tuning. In this work, we propose the Latent Program Network (LPN), a novel architecture that builds in test-time search directly into neural models. LPN learns a latent space of implicit programs--neurally mapping inputs to outputs--through which it can search using gradients at test time.

Graph neural networks (GNNs) can efficiently process text-attributed graphs (TAGs) due to their message-passing mechanisms, but their training heavily relies on the human-annotated labels. Moreover, the complex and diverse local topologies of nodes of real-world TAGs make it challenging for a single mechanism to handle. Large language models (LLMs) perform well in zero-/few-shot learning on TAGs but suffer from a scalability challenge. Therefore, we propose a preference-driven knowledge distillation (PKD) framework to synergize the complementary strengths of LLMs and various GNNs for few-shot node classification. Specifically, we develop a GNN-preference-driven node selector that effectively promotes prediction distillation from LLMs to teacher GNNs. To further tackle nodes' intricate local topologies, we develop a node-preferencedriven GNN selector that identifies the most suitable teacher GNN for each node, thereby facilitating tailored knowledge distillation from teacher GNNs to the student GNN. Extensive experiments validate the efficacy of our proposed framework in few-shot node classification on real-world TAGs.