CoDA: Towards Effective Cross-domain Knowledge Transfer via CoT-guided Domain Adaptation

Large language models (LLMs) have achieved substantial advances in logical reasoning, yet they continue to lag behind human-level performance. In-context learning provides a viable solution that boosts the model's performance via prompting its input with expert-curated, in-domain exemplars. However, in many real-world, expertise-scarce domains, such as low-resource scientific disciplines, emerging biomedical subfields, or niche legal jurisdictions, such high-quality in-domain demonstrations are inherently limited or entirely unavailable, thereby constraining the general applicability of these approaches. To mitigate this limitation, recent efforts have explored the retrieval of cross-domain samples as surrogate in-context demonstrations. Nevertheless, the resulting gains remain modest. This is largely attributable to the pronounced domain shift between source and target distributions, which impedes the model's ability to effectively identify and exploit underlying shared structures or latent reasoning patterns. Consequently, when relying solely on raw textual prompting, LLMs struggle to abstract and transfer such cross-domain knowledge in a robust and systematic manner. To address these issues, we propose CoDA, which employs a lightweight adapter to directly intervene in the intermediate hidden states. By combining feature-based distillation of CoT-enriched reference representations with Maximum Mean Discrepancy (MMD) for kernelized distribution matching, our method aligns the latent reasoning representation of the source and target domains. Extensive experimental results on multiple logical reasoning tasks across various model families validate the efficacy of CoDA by significantly outperforming the previous state-of-the-art baselines by a large margin.

Mitigating Multimodal Hallucination via Phase-wise Self-reward

Large Vision-Language Models (LVLMs) still struggle with vision hallucination, where generated responses are inconsistent with the visual input. Existing methods either rely on large-scale annotated data for fine-tuning, which incurs massive computational overhead, or employ static post-hoc strategies that overlook the dynamic nature of hallucination emergence. To address these, we introduce a new self-rewarding framework, enabling dynamic hallucination mitigation at inference time without external supervision. On the empirical side, we reveal that visual hallucination exhibits phase-wise dynamic patterns, peaking at the onset of each semantic phase. Drawing on these insights, we propose \textbf{PSRD} (\textbf{Phase-wise \textbf{S}elf-\textbf{R}eward \textbf{D}ecoding) for online hallucination correction guided by phase-wise self-reward signals. To reduce the cost of repeated self-evaluation during decoding, we distill the hallucination guidance signal from LVLMs into a lightweight reward model. The reward model subsequently provides on-the-fly guidance for targeted intervention during the decoding process, enabling precise hallucination suppression. The proposed PSRD significantly reduces the hallucination rate of LLaVA-1.5-7B by 50.0% and consistently outperforms existing post-hoc methods across five hallucination evaluation benchmarks for four LVLMs. Further analysis confirms that PSRD effectively mitigates hallucination propagation and achieves a highly controllable trade-off between strong performance and inference efficiency.

Learning step-level dynamic soaring in shear flow

Dynamic soaring enables sustained flight by extracting energy from wind shear, yet it is commonly understood as a cycle-level maneuver that assumes stable flow conditions. In realistic unsteady environments, however, such assumptions are often violated, raising the question of whether explicit cycle-level planning is necessary. Here, we show that dynamic soaring can emerge from step-level, state-feedback control using only local sensing, without explicit trajectory planning. Using deep reinforcement learning as a tool, we obtain policies that achieve robust omnidirectional navigation across diverse shear-flow conditions. The learned behavior organizes into a structured control law that coordinates turning and vertical motion, giving rise to a two-phase strategy governed by a trade-off between energy extraction and directional progress. The resulting policy generalizes across varying conditions and reproduces key features observed in biological flight and optimal-control solutions. These findings identify a feedback-based control structure underlying dynamic soaring, demonstrating that efficient energy-harvesting flight can emerge from local interactions with the flow without explicit planning, and providing insights for biological flight and autonomous systems in complex, flow-coupled environments.

When Is Thinking Enough? Early Exit via Sufficiency Assessment for Efficient Reasoning

Large reasoning models (LRMs) have achieved remarkable performance in complex reasoning tasks, driven by their powerful inference-time scaling capability. However, LRMs often suffer from overthinking, which results in substantial computational redundancy and significantly reduces efficiency. Early-exit methods aim to mitigate this issue by terminating reasoning once sufficient evidence has been generated, yet existing approaches mostly rely on handcrafted or empirical indicators that are unreliable and impractical. In this work, we introduce Dynamic Thought Sufficiency in Reasoning (DTSR), a novel framework for efficient reasoning that enables the model to dynamically assess the sufficiency of its chain-of-thought (CoT) and determine the optimal point for early exit. Inspired by human metacognition, DTSR operates in two stages: (1) Reflection Signal Monitoring, which identifies reflection signals as potential cues for early exit, and (2) Thought Sufficiency Check, which evaluates whether the current CoT is sufficient to derive the final answer. Experimental results on the Qwen3 models show that DTSR reduces reasoning length by 28.9%-34.9% with minimal performance loss, effectively mitigating overthinking. We further discuss overconfidence in LRMs and self-evaluation paradigms, providing valuable insights for early-exit reasoning.

Towards Effective In-context Cross-domain Knowledge Transfer via Domain-invariant-neurons-based Retrieval

Large language models (LLMs) have made notable progress in logical reasoning, yet still fall short of human-level performance. Current boosting strategies rely on expert-crafted in-domain demonstrations, limiting their applicability in expertise-scarce domains, such as specialized mathematical reasoning, formal logic, or legal analysis. In this work, we demonstrate the feasibility of leveraging cross-domain demonstrating examples to boost the LLMs' reasoning performance. Despite substantial domain differences, many reusable implicit logical structures are shared across domains. In order to effectively retrieve cross-domain examples for unseen domains under investigation, in this work, we further propose an effective retrieval method, called domain-invariant neurons-based retrieval (\textbf{DIN-Retrieval}). Concisely, DIN-Retrieval first summarizes a hidden representation that is universal across different domains. Then, during the inference stage, we use the DIN vector to retrieve structurally compatible cross-domain demonstrations for the in-context learning. Experimental results in multiple settings for the transfer of mathematical and logical reasoning demonstrate that our method achieves an average improvement of 1.8 over the state-of-the-art methods \footnote{Our implementation is available at https://github.com/Leon221220/DIN-Retrieval}.

Reason Analogically via Cross-domain Prior Knowledge: An Empirical Study of Cross-domain Knowledge Transfer for In-Context Learning

Despite its success, existing in-context learning (ICL) relies on in-domain expert demonstrations, limiting its applicability when expert annotations are scarce. We posit that different domains may share underlying reasoning structures, enabling source-domain demonstrations to improve target-domain inference despite semantic mismatch. To test this hypothesis, we conduct a comprehensive empirical study of different retrieval methods to validate the feasibility of achieving cross-domain knowledge transfer under the in-context learning setting. Our results demonstrate conditional positive transfer in cross-domain ICL. We identify a clear example absorption threshold: beyond it, positive transfer becomes more likely, and additional demonstrations yield larger gains. Further analysis suggests that these gains stem from reasoning structure repair by retrieved cross-domain examples, rather than semantic cues. Overall, our study validates the feasibility of leveraging cross-domain knowledge transfer to improve cross-domain ICL performance, motivating the community to explore designing more effective retrieval approaches for this novel direction.\footnote{Our implementation is available at https://github.com/littlelaska/ICL-TF4LR}

Tac2Real: Reliable and GPU Visuotactile Simulation for Online Reinforcement Learning and Zero-Shot Real-World Deployment

Visuotactile sensors are indispensable for contact-rich robotic manipulation tasks. However, policy learning with tactile feedback in simulation, especially for online reinforcement learning (RL), remains a critical challenge, as it demands a delicate balance between physics fidelity and computational efficiency. To address this challenge, we present Tac2Real, a lightweight visuotactile simulation framework designed to enable efficient online RL training. Tac2Real integrates the Preconditioned Nonlinear Conjugate Gradient Incremental Potential Contact (PNCG-IPC) method with a multi-node, multi-GPU high-throughput parallel simulation architecture, which can generate marker displacement fields at interactive rates. Meanwhile, we propose a systematic approach, TacAlign, to narrow both structured and stochastic sources of domain gap, ensuring a reliable zero-shot sim-to-real transfer. We further evaluate Tac2Real on the contact-rich peg insertion task. The zero-shot transfer results achieve a high success rate in the real-world scenario, verifying the effectiveness and robustness of our framework. The project page is: https://ningyurichard.github.io/tac2real-project-page/

Mitigating Translationese Bias in Multilingual LLM-as-a-Judge via Disentangled Information Bottleneck

Large language models (LLMs) have become a standard for multilingual evaluation, yet they exhibit a severe systematic translationese bias. In this paper, translationese bias is characterized as LLMs systematically favoring machine-translated text over human-authored references, particularly in low-resource languages. We attribute this bias to spurious correlations with (i) latent manifold alignment with English and (ii) cross-lingual predictability. To mitigate this bias, we propose DIBJudge, a robust fine-tuning framework that learns a minimally sufficient, judgment-critical representation via variational information compression, while explicitly isolating spurious factors into the dedicated bias branch. Furthermore, we incorporate a cross-covariance penalty that explicitly suppresses statistical dependence between robust and bias representations, thereby encouraging effective disentanglement. Extensive evaluations on multilingual reward modeling benchmarks and a dedicated translationese bias evaluation suite demonstrate that the proposed DIBJudge consistently outperforms strong baselines and substantially mitigates translationese bias.

Timer-S1: A Billion-Scale Time Series Foundation Model with Serial Scaling

We introduce Timer-S1, a strong Mixture-of-Experts (MoE) time series foundation model with 8.3B total parameters, 0.75B activated parameters for each token, and a context length of 11.5K. To overcome the scalability bottleneck in existing pre-trained time series foundation models, we perform Serial Scaling in three dimensions: model architecture, dataset, and training pipeline. Timer-S1 integrates sparse TimeMoE blocks and generic TimeSTP blocks for Serial-Token Prediction (STP), a generic training objective that adheres to the serial nature of forecasting. The proposed paradigm introduces serial computations to improve long-term predictions while avoiding costly rolling-style inference and pronounced error accumulation in the standard next-token prediction. Pursuing a high-quality and unbiased training dataset, we curate TimeBench, a corpus with one trillion time points, and apply meticulous data augmentation to mitigate predictive bias. We further pioneer a post-training stage, including continued pre-training and long-context extension, to enhance short-term and long-context performance. Evaluated on the large-scale GIFT-Eval leaderboard, Timer-S1 achieves state-of-the-art forecasting performance, attaining the best MASE and CRPS scores as a pre-trained model. Timer-S1 will be released to facilitate further research.

TextBoost: Boosting Scene Text Fidelity in Ultra-low Bitrate Image Compression

Ultra-low bitrate image compression faces a critical challenge: preserving small-font scene text while maintaining overall visual quality. Region-of-interest (ROI) bit allocation can prioritize text but often degrades global fidelity, leading to a trade-off between local accuracy and overall image quality. Instead of relying on ROI coding, we incorporate auxiliary textual information extracted by OCR and transmitted with negligible overhead, enabling the decoder to leverage this semantic guidance. Our method, TextBoost, operationalizes this idea through three strategic designs: (i) adaptively filtering OCR outputs and rendering them into a guidance map; (ii) integrating this guidance with decoder features in a calibrated manner via an attention-guided fusion block; and (iii) enforcing guidance-consistent reconstruction in text regions with a regularizing loss that promotes natural blending with the scene. Extensive experiments on TextOCR and ICDAR 2015 demonstrate that TextBoost yields up to 60.6% higher text-recognition F1 at comparable Peak Signal-to-Noise Ratio (PSNR) and bits per pixel (bpp), producing sharper small-font text while preserving global image quality and effectively decoupling text enhancement from global rate-distortion optimization.