3D Gaussian Splatting (3D-GS) has emerged as an efficient 3D representation and a promising foundation for semantic tasks like segmentation. However, existing 3D-GS-based segmentation methods typically rely on high-dimensional category features, which introduce substantial memory overhead. Moreover, fine-grained segmentation remains challenging due to label space congestion and the lack of stable multi-granularity control mechanisms. To address these limitations, we propose a coarse-to-fine binary encoding scheme for per-Gaussian category representation, which compresses each feature into a single integer via the binary-to-decimal mapping, drastically reducing memory usage. We further design a progressive training strategy that decomposes panoptic segmentation into a series of independent sub-tasks, reducing inter-class conflicts and thereby enhancing fine-grained segmentation capability. Additionally, we fine-tune opacity during segmentation training to address the incompatibility between photometric rendering and semantic segmentation, which often leads to foreground-background confusion. Extensive experiments on multiple benchmarks demonstrate that our method achieves state-of-the-art segmentation performance while significantly reducing memory consumption and accelerating inference.

Yuzhen Shi, Huanghai Liu, Yiran HU, Song Gaojie, Xu Xinran, Yubo Ma, Tianyi Tang, Li Zhang, Qingjing Chen, Feng Di, Wenbo Lv, Weiheng Wu, Kexin Yang, Sen Yang, Wei Wang, Rongyao Shi, Qiu Yuanyang, Yuemeng Qi, Zhang Jingwen, Sui Xiaoyu, Yifan Chen, Zhang Yi, An Yang, Bowen Yu, Dayiheng Liu, Junyang Lin, Weixing Shen, Bing Zhao, Charles L. A. Clarke, HU Wei. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a powerful approach to enhancing the reasoning capabilities of Large Language Models (LLMs), yet its underlying mechanisms remain insufficiently understood. In this work, we undertake a pioneering exploration of RLVR through the novel perspective of token entropy patterns, comprehensively analyzing how different tokens influence reasoning performance. By examining token entropy patterns in Chain-of-Thought (CoT) reasoning, we observe that only a small fraction (approximately 20\%) of tokens exhibit high entropy, and these tokens semantically act as critical forks that steer the model toward diverse reasoning pathways. We further demonstrate that moderately increasing the entropy of these high-entropy tokens via decoding temperature adjustments leads to improved performance, quantitatively confirming their role as decision points in reasoning. We ultimately refine RLVR by restricting policy gradient updates to these forking tokens. Despite utilizing only 20\% of tokens, our approach achieves comparable performance to full-gradient updates on the Qwen3-8B base model. Moreover, it demonstrates remarkable improvements on the larger Qwen3-32B base model, boosting AIME'25 scores by 11.04 and AIME'24 scores by 7.71. In contrast, training exclusively on the 80\% lowest-entropy tokens leads to a marked decline in performance. These findings indicate that the efficacy of RLVR primarily arises from optimizing the high-entropy tokens that dictate key reasoning directions. Collectively, our results suggest promising avenues for optimizing RLVR algorithms by strategically leveraging the potential of these high-entropy minority tokens to further enhance the reasoning abilities of LLMs.

Neural architecture search (NAS) has shown great promise in automatically designing neural network models. Recently, block-wise NAS has been proposed to alleviate deep coupling problem between architectures and weights existed in the well-known weight-sharing NAS, by training the huge weight-sharing supernet block-wisely. However, the existing block-wise NAS methods, which resort to either supervised distillation or self-supervised contrastive learning scheme to enable block-wise optimization, take massive computational cost. To be specific, the former introduces an external high-capacity teacher model, while the latter involves supernet-scale momentum model and requires a long training schedule. Considering this, in this work, we propose a resource-friendly deeply supervised block-wise NAS (DBNAS) method. In the proposed DBNAS, we construct a lightweight deeply-supervised module after each block to enable a simple supervised learning scheme and leverage ground-truth labels to indirectly supervise optimization of each block progressively. Besides, the deeply-supervised module is specifically designed as structural and functional condensation of the supernet, which establishes global awareness for progressive block-wise optimization and helps search for promising architectures. Experimental results show that the DBNAS method only takes less than 1 GPU day to search out promising architectures on the ImageNet dataset with less GPU memory footprint than the other block-wise NAS works. The best-performing model among the searched DBNAS family achieves 75.6% Top-1 accuracy on ImageNet, which is competitive with the state-of-the-art NAS models. Moreover, our DBNAS family models also achieve good transfer performance on CIFAR-10/100, as well as two downstream tasks: object detection and semantic segmentation.

Although resistance monitoring write termination (RM-WT) scheme for STT-MRAM can reduce the write energy, the degradation of read margin due to low tunnel magnetoresistance ratio (TMR) and intrusion of noise with process variation still seriously deteriorates the stability of the WT operation. In this paper, a double-ended superposition anti-noise write termination (DSA-WT) scheme is proposed and implemented, in which the voltage changes on both BL and SL can be superimposed to boost sensing margin (SM). Schmitt trigger (ST) is adopted to take the place of the inverter (INV) in the traditional WT scheme, which is demonstrated to be helpful for stability improvement. Based on 65-nm CMOS technology, the proposed DSA-WT scheme shows 15% ~33% sensing margin boosting under various PVT conditions and 1000 times lower read bit-error-rate (BER) compared with the other WT schemes. The write done (WD) delay and the energy-delay-product (EDP) achieve 49.6% and 47.2% improvements compared to the state-of-art self-referenced single-ended RM-WT scheme (SS-RM-WT), respectively.

In this work, we pursue a unified paradigm for multimodal pretraining to break the shackles of complex task/modality-specific customization. We propose OFA, a Task-Agnostic and Modality-Agnostic framework that supports Task Comprehensiveness. OFA unifies a diverse set of cross-modal and unimodal tasks, including image generation, visual grounding, image captioning, image classification, language modeling, etc., in a simple sequence-to-sequence learning framework. OFA follows the instruction-based learning in both pretraining and finetuning stages, requiring no extra task-specific layers for downstream tasks. In comparison with the recent state-of-the-art vision & language models that rely on extremely large cross-modal datasets, OFA is pretrained on only 20M publicly available image-text pairs. Despite its simplicity and relatively small-scale training data, OFA achieves new SOTAs in a series of cross-modal tasks while attaining highly competitive performances on uni-modal tasks. Our further analysis indicates that OFA can also effectively transfer to unseen tasks and unseen domains. Our code and models are publicly available at https://github.com/OFA-Sys/OFA.

Shuhuai Ren, Junyang Lin, Guangxiang Zhao, Rui Men, An Yang, Jingren Zhou, Xu Sun, Hongxia Yang. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Multimodal pretraining has demonstrated success in the downstream tasks of cross-modal representation learning. However, it is limited to the English data, and there is still a lack of large-scale dataset for multimodal pretraining in Chinese. In this work, we propose the largest dataset for pretraining in Chinese, which consists of over 1.9TB images and 292GB texts. The dataset has large coverage over domains, including encyclopedia, question answering, forum discussion, etc. Besides, we propose a method called M6, referring to Multi-Modality-to-Multi-Modality Multitask Mega-transformer, for unified pretraining on the data of single modality and multiple modalities. The model is pretrained with our proposed tasks, including text-to-text transfer, image-to-text transfer, as well as multi-modality-to-text transfer. The tasks endow the model with strong capability of understanding and generation. We scale the model to 10 billion parameters, and build the largest pretrained model in Chinese. Experimental results show that our proposed M6 outperforms the baseline in a number of downstream tasks concerning both single modality and multiple modalities, and the 10B-parameter pretrained model demonstrates strong potential in the setting of zero-shot learning.

Lacking robustness is a serious problem for Machine Reading Comprehension (MRC) models. To alleviate this problem, one of the most promising ways is to augment the training dataset with sophisticated designed adversarial examples. Generally, those examples are created by rules according to the observed patterns of successful adversarial attacks. Since the types of adversarial examples are innumerable, it is not adequate to manually design and enrich training data to defend against all types of adversarial attacks. In this paper, we propose a novel robust adversarial training approach to improve the robustness of MRC models in a more generic way. Given an MRC model well-trained on the original dataset, our approach dynamically generates adversarial examples based on the parameters of current model and further trains the model by using the generated examples in an iterative schedule. When applied to the state-of-the-art MRC models, including QANET, BERT and ERNIE2.0, our approach obtains significant and comprehensive improvements on 5 adversarial datasets constructed in different ways, without sacrificing the performance on the original SQuAD development set. Moreover, when coupled with other data augmentation strategy, our approach further boosts the overall performance on adversarial datasets and outperforms the state-of-the-art methods.

Machine reading comprehension (MRC) is a crucial and challenging task in NLP.Recently, pre-trained language models (LMs), especially BERT, have achieved remarkable success, presenting new state-of-the-art results in MRC.In this work, we investigate the potential of leveraging external knowledge bases (KBs) to further improve BERT for MRC.We introduce KT-NET, which employs an attention mechanism to adaptively select desired knowledge from KBs, and then fuses selected knowledge with BERT to enable context-and knowledgeaware predictions.We believe this would combine the merits of both deep LMs and curated KBs towards better MRC.Experimental results indicate that KT-NET offers significant and consistent improvements over BERT, outperforming competitive baselines on ReCoRD and SQuAD1.1 benchmarks.Notably, it ranks the 1st place on the ReCoRD leaderboard, and is also the best single model on the SQuAD1.1 leaderboard at the time of submission (March 4th, 2019). 1

The security of Industrial Control Systems (ICS) is closely related to national security. With secret exploration and analysis of a target ICS, highly-skilled attackers can gain enough key knowledge about the system (e.g., the physical model of the system and the corresponding detection threshold), and then launch stealthy attacks by keeping the detection indicator under its threshold, thus bypasses existing intrusion detection mechanisms. However, we discover that all devices in industrial control systems consume energy at run time and the energy consumption varies according to different operation types and system states. Therefore, there exists relationships between control operation, system state and energy consumption of the device. Accordingly, we put forward a novel ICS intrusion detection approach based on multi-dimensional data fusion. This approach collects information about power consumption of physical devices, control operation and system state, and then identifies stealthy attacks by feeding the multi-dimensional information into a cascade detection algorithm. Experimental results verify that our approach has a better detection performance than other detection methods.

The Industrial Control System (ICS) refers to the national critical infrastructure, such as Energy and Water facility, which is significant for the national security. Sequence attack is a unique attack type in ICS, and many detection approaches have been proposed. A common and unrealistic hypothesis of these approaches is that they have gained the command sequences. In the real world, we can only obtain the observations from sensors. The single observation detection technique is a common approach to find anomalies by the observations. However, the highly skilled attacker can compromise some Programmable Logic Controllers (PLCs) in ICS and fake their sensor measurements. Under this circumstance, this detection approach becomes invalid and increases the false-negative rate. In this paper, we first analyze the sequence attack by their attack capability in ICS. Then we propose a State-Based Sequence Detection approach (SBSD). The SBSD uses the equipment's observation information, belonging to many PLCs, to create Hidden Markov Models (HMMs) for detecting the sequence attack. The experiment results in an ICS testbed have shown the effectiveness of SBSD.

Beimel and Orlov proved that all information inequalities on four or five variables, together with all information inequalities on more than five variables that are known to date, provide lower bounds on the size of the shares in secret sharing schemes that are at most linear on the number of participants. We present here another two negative results about the power of information inequalities in the search for lower bounds in secret sharing. First, we prove that all information inequalities on a bounded number of variables can only provide lower bounds that are polynomial on the number of participants. Second, we prove that the rank inequalities that are derived from the existence of two common informations can provide only lower bounds that are at most cubic in the number of participants.

We present new families of access structures that, similarly to the multilevel and compartmented access structures introduced in previous works, are natural generalizations of threshold secret sharing. Namely, they admit ideal linear secret sharing schemes over every large enough finite field, they can be described by a small number of parameters, and they have useful properties for the applications of secret sharing. The use of integer polymatroids makes it possible to find many new such families and it simplifies in great measure the proofs for the existence of ideal secret sharing schemes for them.

In 1986, D. V. Chudnovsky and G. V. Chudnovsky first employed algebraic curves over finite fields to construct bilinear multiplication algorithms implicitly through supercodes introduced by Shparlinski-Tsfasman-Vladuţ, or equivalently, multiplication-friendly codes that we will introduce in this paper. This idea was further developed by Shparlinski-Tsfasman-Vladuţ in order to study the asymptotic behavior of multiplication complexity in extension fields. Later on, Ballet et al. further investigated the method and obtained some improvements. Recently, Ballet and Pieltant made use of curves over an extension field of to obtain an improvement on the complexity of multiplications in extensions of the binary field. In this paper, we develop the multiplication-friendly splitting technique and then apply this technique to study asymptotic behavior of multiplications in extension fields. By combining this with the idea of using algebraic function fields, we are able to improve further the asymptotic results of multiplication complexity. In particular, the improvement for small fields such as the binary and ternary fields is substantial.