The Mixture of Experts (MoE) architecture is a cornerstone of modern state-of-the-art (SOTA) large language models (LLMs). MoE models facilitate scalability by enabling sparse parameter activation. However, traditional MoE architecture uses homogeneous experts of a uniform size, activating a fixed number of parameters irrespective of input complexity and thus limiting computational efficiency. To overcome this limitation, we introduce Grove MoE, a novel architecture incorporating experts of varying sizes, inspired by the heterogeneous big.LITTLE CPU architecture. This architecture features novel adjugate experts with a dynamic activation mechanism, enabling model capacity expansion while maintaining manageable computational overhead. Building on this architecture, we present GroveMoE-Base and GroveMoE-Inst, 33B-parameter LLMs developed by applying an upcycling strategy to the Qwen3-30B-A3B-Base model during mid-training and post-training. GroveMoE models dynamically activate 3.14 - 3.28B parameters based on token complexity and achieve performance comparable to SOTA open-source models of similar or even larger size.

Diffusion large language models (dLLMs) generate text through iterative denoising. In commonly adopted parallel decoding schemes, each step confirms only high-confidence positions while remasking the others. By analyzing dLLM denoising traces, we uncover a key inefficiency: models often predict the correct target token several steps before its confidence becomes high enough to be decoded. This gap between early prediction and late decoding forces repeated remasking of already-correct tokens, causing redundant iterations and limiting acceleration. To exploit this temporal redundancy, we introduce Trace Credit to quantify a token’s decoding potential by accumulating historical evidence. Building on this, we propose CreditDecoding, a training-free parallel decoding method that fuses Trace Credit with current logits to boost the confidence of correct but underconfident tokens, thereby accelerating denoising and improving robustness. On eight benchmarks, CreditDecoding achieves up to 5.48 times speedup with +0.48 accuracy on LLaDA-8B and consistently improves performance across diverse dLLM architectures and parameter scales. It further scales to long contexts and remains orthogonal to mainstream inference optimizations, making it a practical and widely applicable solution.

Deep semantic retrieval has achieved remarkable success in online E-commerce applications. The majority of methods aim to distinguish positive items and negative items for each query by utilizing margin loss or softmax loss. Despite their decent performance, these methods are highly sensitive to hyper-parameters, i.e., margin and temperature 𝜏, which measure the similarity of negative pairs and affect the distribution of items in metric space. How to design and choose adaptively parameters for different pairs is still an open challenge. Recently several methods have attempted to alleviate the above problem by learning each parameter through trainable/statistical methods in the recommendation. We argue that those are not suitable for retrieval scenarios, due to the agnosticism and diversity of the queries. To fully overcome this limitation, we propose a novel adaptive metric learning method that designs a simple and universal hyper-parameter-free learning method to improve the performance of retrieval. Specifically, we first propose a method that adaptive obtains the hyper-parameters by relying on the batch similarity without fixed or extra-trainable hyper-parameters. Subsequently, we adopt a symmetric metric learning method to mitigate model collapse issues. Furthermore, the proposed method is general and sheds a highlight on other fields. Extensive experiments demonstrate our method significantly outperforms previous methods on a real-world dataset, highlighting the superiority and effectiveness of our method. This method has been successfully deployed on an online E-commerce search platform and brought substantial economic benefits.

To allow an easy understanding of the various licenses that exist for the use of Language Resources (ELRA’s, META-SHARE’s, Creative Commons’, etc.), ELRA has developed a License Wizardto help the right-holders share/distribute their resources under the appropriate license. It also aims to be exploited by users to better understand the legal obligations that apply in various licensing situations. The present paper elaborates on the License Wizard functionalities of this web configurator, which enables to select a number of legal features and obtain the user license adapted to the users selection, to define which user licenses they would like to select in order to distribute their Language Resources, to integrate the user license terms into a Distribution Agreement that could be proposed to ELRA or META-SHARE for further distribution through the ELRA Catalogue of Language Resources. Thanks to a flexible back office, the structure of the legal feature selection can easily be reviewed to include other features that may be relevant for other licenses. Integrating contributions from other initiatives thus aim to be one of the obvious next steps, with a special focus on CLARIN and Linked Data experiences.

In this paper we describe the new developments brought to LRE Map, especially in terms of the user interface of the Web application, of the searching of the information therein, and of the data model updates.

This article presents the latest dissemination activities and technical developments that were carried out for the International Standard Language Resource Number (ISLRN) service. It also recalls the main principle and submission process for providers to obtain their 13-digit ISLRN identifier. Up to March 2016, 2100 Language Resources were allocated an ISLRN number, not only ELRA’s and LDC’s catalogued Language Resources, but also the ones from other important organisations like the Joint Research Centre (JRC) and the Resource Management Agency (RMA) who expressed their strong support to this initiative. In the research field, not only assigning a unique identification number is important, but also referring to a Language Resource as an object per se (like publications) has now become an obvious requirement. The ISLRN could also become an important parameter to be considered to compute a Language Resource Impact Factor (LRIF) in order to recognize the merits of the producers of Language Resources. Integrating the ISLRN number into a LR-oriented bibliographical reference is thus part of the objective. The idea is to make use of a BibTeX entry that would take into account Language Resources items, including ISLRN.The ISLRN being a requested field within the LREC 2016 submission, we expect that several other LRs will be allocated an ISLRN number by the conference date. With this expansion, this number aims to be a spreadly-used LR citation instrument within works referring to LRs.