Tables present unique challenges for language models due to their structured row-column interactions, necessitating specialized approaches for effective comprehension. While large language models (LLMs) have demonstrated potential in table reasoning through prompting and techniques like chain-of-thought (CoT) and program-of-thought (PoT), optimizing their performance for table question answering remains underexplored. In this paper, we introduce region-based Table-R1, a novel reinforcement learning approach that enhances LLM table understanding by integrating region evidence into reasoning steps. Our method employs Region-Enhanced Supervised Fine-Tuning (RE-SFT) to guide models in identifying relevant table regions before generating answers, incorporating textual, symbolic, and program-based reasoning. Additionally, Table-Aware Group Relative Policy Optimization (TARPO) introduces a mixed reward system to dynamically balance region accuracy and answer correctness, with decaying region rewards and consistency penalties to align reasoning steps. Experiments show that Table-R1 achieves an average performance improvement of 14.36 points across multiple base models on three benchmark datasets, even outperforming baseline models with ten times the parameters, while TARPO reduces response token consumption by 67.5

In smart cities composed of numerous IoT devices, multi-party collaboration for data sharing and intelligent prediction plays a crucial role in enabling precise decision-making. The data involved in the intelligent prediction process often contains large amounts of sensitive information. Therefore, ensuring information security is of paramount importance. Homomorphic encryption (HE) enables computation without revealing the original data, ensuring data privacy during secure multi-party computation (MPC). This enables HE to ensure information security for multi-party collaborative computation when dealing with sensitive data. Compared to traditional single-key HE, multi-key fully HE (MK-FHE) better meets the needs of multi-party cooperation, enhancing both security and flexibility. For MK-FHE schemes that can achieve intelligent prediction, the computational complexity of existing MK-FHE schemes increases at least quadratically with the number of keys during multiplication operations. We propose an optimized MK-FHE multiplication algorithm that reduces the computational complexity from quadratic to linear through an innovative relinearization algorithm, achieving optimal efficiency asymptotically. Subsequently, we design an intelligent prediction protocol for a multi-party collaboration scenario, and apply the optimized MK-FHE scheme to it. Experimental results show that, In a typical IoT scenario simulation, our scheme has a significant advantage in terms of time consumption. Specifically, compared to the previous intelligent prediction scheme employing MK-FHE that had the lowest computational complexity, with 72 keys, multiple key CKKS (MK-CKKS) achieves up to an 8.6× speedup, and multiple key BFV (MK-BFV) achieves up to an 11.3× speedup. Additionally, it performs well in noise control and neural network (NN) model performance (including F1-score, recall, precision, and accuracy).

Multimodal Sentiment Analysis (MSA) integrates multiple modalities to infer human sentiment, but real-world noise often leads to missing or corrupted data. However, existing feature-disentangled methods struggle to handle the internal variations of heterogeneous information under uncertain missingness, making it difficult to learn effective multimodal representations from degraded modalities. To address this issue, we propose DERL, a Disentangled Expert Representation Learning framework for robust MSA. Specifically, DERL employs hybrid experts to adaptively disentangle multimodal inputs into orthogonal private and shared representation spaces. A multi-level reconstruction strategy is further developed to provide collaborative supervision, enhancing both the expressiveness and robustness of the learned representations. Finally, the disentangled features act as modality experts with distinct roles to generate importance-aware fusion results. Extensive experiments on three MSA benchmarks demonstrate that DERL outperforms state-of-the-art methods under various missing-modality conditions. For instance, our method improves Acc-2 by 2.22% and reduces MAE by 2.25% on CMU-MOSI under intra-modal missingness.

Large language models (LLMs) significantly enhance their ability to process tabular data through chain-of-thought reasoning, particularly in table question answering tasks. However, LLMs encounter substantial challenges when dealing with large tables in real-world applications. Prompting LLMs with the entire table not only encounters context-length constraints but also significantly extends the reasoning path, heightening the risk of reasoning hallucination and information truncation. To address this, we construct a large-size table reasoning (LSTR) benchmark, featuring tables larger than those in existing benchmarks, to thoroughly investigate how table size affects the reasoning abilities of LLMs in answering table-related questions. Subsequently, we propose a size-adaptive-thought (SAT) approach that instructs the LLM utilizing refined metadata to employ Python commands for manipulating tables step by step, thereby facilitating efficient reasoning with tables of any size. Furthermore, we develop SAT-Llama, fine-tuned SAT on Llama3.1 (8B), which delivers performance comparable to large-size LLMs at a much lower cost, addressing the issue of inadequate code manipulation capabilities in small-size LLMs. Experimental results on the LSTR and WTQ datasets demonstrate that SAT achieves a new state-of-the-art in handling large-size tables, exhibiting significant performance advantages and high context token efficiency.

Zero-shot tweet-level stance detection confronts two primary challenges: (1) mitigating the context sparsity inherent in short texts, and (2) establishing the relevance between implicit targets and textual content. While existing methods primarily focus on incorporating external knowledge, they neglect the intrinsic semantic cues embedded within key intra-textual entities. Furthermore, current models exhibit limited capability in determining the relevance of unseen targets to the given text, thereby struggling to differentiate between "neutral" and "irrelevant" stance labels. To address these issues, we first construct a four-class, multi-topic Japanese tweet dataset. To our knowledge, this is the first Japanese tweet-level dataset for stance detection. We then propose KIRP, a zero-shot stance detection framework. It integrates external knowledge with entity reorganization for data augmentation and employs prompt chaining for reasoning. Specifically, the framework incorporates knowledge graphs to supplement and reorganize key textual entities, while reflective Chain-of-Thought (CoT) reasoning extracts and validates implicit targets. To better distinguish "neutral" from "irrelevant" labels, we adopt stance-aware contrastive learning to capture discriminative features and design a three-layer iterative prototype network for fine-grained classification. Experimental results on SemEval-2016, WT-WT, and KIRP-D show that KIRP achieves state-of-the-art performance. KIRP obtains F1 scores of 84.05

In modern software development, particularly in emerging ``vibe coding'' paradigms, project implementation increasingly begins with visual interactions between users and AI coding assistants, where system architectures are communicated through visual designs before coding. This visual-first approach necessitates AI systems capable of interpreting diagrams across multiple programming languages. However, the development of such systems is severely hindered by the lack of large-scale multimodal training data and evaluation benchmarks. To address these limitations, we present M2C-INSTRUCT, a comprehensive multilingual multimodal instruction-tuning dataset containing over 13.1M samples across 50+ programming languages, designed for visual understanding and diagram interpretation in code generation tasks. We validate our dataset by training M2-CODER, a multilingual multimodal software developer that successfully integrates visual design inputs with textual instructions. We also introduce M2EVAL, a novel multilingual evaluation benchmark for multimodal code generation performance. Experiments show our 7B M2-CODER performs on par with much larger 70B+ models, confirming the quality and effectiveness of our M2C-INSTRUCT. Together, M2C-INSTRUCT, M2-CODER, and M2EVAL provide essential infrastructure for visual-assisted programming in vibe-coding and visual-interactive development workflows.

Industrial software development across chip design, GPU optimization, and embedded systems lacks expert reasoning traces showing how engineers reason about hardware constraints and timing semantics. In this work, we propose InCoder-32B-Thinking, trained on the data from the Error-driven Chain-of-Thought (ECoT) synthesis framework with an industrial code world model (ICWM) to generate reasoning traces. Specifically, ECoT generates reasoning chains by synthesizing the thinking content from multi-turn dialogue with environmental error feedback, explicitly modeling the error-correction process. ICWM is trained on domain-specific execution traces from Verilog simulation, GPU profiling, etc., learns the causal dynamics of how code affects hardware behavior, and enables self-verification by predicting execution outcomes before actual compilation. All synthesized reasoning traces are validated through domain toolchains, creating training data matching the natural reasoning depth distribution of industrial tasks. Evaluation on 14 general (81.3

Code large language models (LLMs) have made significant progress in code debugging by directly generating the correct code based on the buggy code snippet. Programming benchmarks, typically consisting of buggy code snippet and their associated test cases, are used to assess the debugging capabilities of LLMs. However, many existing benchmarks primarily focus on Python and are often limited in terms of language diversity (e.g., DebugBench and DebugEval). To advance the field of multilingual debugging with LLMs, we propose the first massively multilingual debugging benchmark, which includes 3.6K test samples of 18 programming languages and covers the automated program repair (APR) task, the code review (CR) task, and the bug identification (BI) task. Further, we introduce the debugging instruction corpora MDEVAL-INSTRUCT by injecting bugs into the correct multilingual queries and solutions (xDebugGen). Further, a multilingual debugger xDebugCoder trained on MDEVAL-INSTRUCT as a strong baseline specifically to handle the bugs of a wide range of programming languages (e.g. "Missing Mut" in language Rust and "Misused Macro Definition" in language C). Our extensive experiments on MDEVAL reveal a notable performance gap between open-source models and closed-source LLMs (e.g., GPT and Claude series), highlighting huge room for improvement in multilingual code debugging scenarios.

Multimodal sentiment analysis (MSA) has attracted increasing attention for its ability to exploit complementary emotional cues from multiple modalities. However, existing methods still encounter two critical limitations:(1) Overemphasis on cross-modal alignment while neglecting in-depth analysis of emotion-specific cross-modal interaction cues, and (2) Reliance on limited labeled data, leading to overfitting in supervised models. To address these challenges, this paper proposes SKAN: A Self-supervised Knowledge-Augmented Network for Multimodal Sentiment Analysis. First, multimodal information is input to a large vision-language model to generate explicit cross-modal sentiment descriptions. The sentiment descriptions, acting as external knowledge, are integrated with the corresponding text-image pairs through a text-centric multimodal fusion module. It augments the model’s ability to discover latent sentiment correlations and improves multimodal sentiment expression capabilities. Second, to alleviate the impact of data scarcity, a self-supervised pretraining strategy is devised, leveraging a sentiment intensity lexicon to perform emotion masking and intensity estimation on unlabeled multimodal data. This design enables the model to acquire cross-modal emotional representations from vast unlabeled samples, thereby improving its semantic sensitivity and generalization ability. Extensive experiments on three benchmark datasets validate the superior performance of SKAN compared with state-of-the-art baselines. The proposed framework provides a novel paradigm that synergistically integrates external knowledge and self-supervision to advance the field of multimodal sentiment analysis.

Large Language Models (LLMs) are constrained by their lack of a long-term memory mechanism, which hinders their ability to maintain context over extended periods and leads to the loss of crucial historical information. To address this limitation, in this paper, we propose the Self-Controlled Memory (SCM) framework to enhance the ability of LLMs to maintain long-term memory and recall relevant information. Our SCM framework comprises three key components: an LLM-based agent serving as the backbone of the framework, a memory stream storing agent memories, and a memory controller updating memories and determining when and how to use the memories from the memory stream. Furthermore, we annotate a dataset, MemoEval, to assess the efficiency of SCM in utilizing memories and processing lengthy inputs. The MemoEval dataset covers three tasks: long-term dialogues, book summarization, and meeting summarization. Experimental results reveal that our SCM framework significantly increases overall accuracy by about 40 https://github.com/wbbeyourself/SCM4LLMs ).

Academic posters are vital for scholarly communication, yet their manual creation is time-consuming. However, automated academic poster generation faces significant challenges in preserving intricate scientific details and achieving effective visual-textual integration. Existing approaches often struggle with semantic richness and structural nuances, and lack standardized benchmarks for evaluating generated academic posters comprehensively. To address these limitations, we introduce P2P, the first flexible, LLM-based multi-agent framework that generates high-quality, HTML-rendered academic posters directly from research papers, demonstrating strong potential for practical applications. P2P employs three specialized agents-for visual element processing, content generation, and final poster assembly-each integrated with dedicated checker modules to enable iterative refinement and ensure output quality. To foster advancements and rigorous evaluation in this domain, we construct and release P2PInstruct, the first large-scale instruction dataset comprising over 30,000 high-quality examples tailored for the academic paper-to-poster generation task. Furthermore, we establish P2PEval, a comprehensive benchmark featuring 121 paper-poster pairs and a dual evaluation methodology (Universal and Fine-Grained) that leverages LLM-as-a-Judge and detailed, human-annotated checklists. Our contributions aim to streamline research dissemination and provide the community with robust tools for developing and evaluating next-generation poster generation systems.

World Action Models (WAMs) unify future environment prediction with action generation for autonomous driving, yet existing approaches optimize only the final outputs, leaving intermediate representations as incidental byproducts. We present ReWorld, the first representation learning framework specifically designed for autonomous-driving WAMs. ReWorld explicitly optimizes the latent world-to-action pathway through three complementary mechanisms. First, it imposes future-predictive supervision on intermediate Video DiT states to encode temporal scene dynamics, enabling self-guided sampling and a roughly twofold convergence speedup. Second, it aligns Action DiT states with their attended video readouts so that the retrieved world information is retained in the representations used for planning. Third, it shapes the action space using geometrically close yet low-scoring hard negatives to separate the expert trajectory from nearby unsafe alternatives. ReWorld constructs supervision entirely from the WAM's own generation targets and attended features, requiring no external encoders or teacher models and introducing only 0.3

Agentic Reasoning has become a transformative force in financial analysis due to its ability to integrate large-scale information and generate reliable and accurate content. However, when handling complex real-world problems, different agents still show significant performance variation. In this work, we design Finance-LaTeX SKILL, a skill for synthesizing financial documents with complex layouts based on expert knowledge. Using an agent workflow built on this skill, we generate 2,000 professional financial documents along with 6,000 high-quality question-answer pairs. To evaluate the overall capability of agents, we introduce FinanceComplexQA, a comprehensive open-ended generation benchmark for financial documents that closely resembles real-world scenarios. It contains 2,026 deep research tasks targeting 1009 financial documents. FinanceComplexQA has 8 key features: bilingual support; coverage of six mainstream scenarios and seven tasks; expert-level document reasoning questions; deep research of complex layouts; relatively stable and permanent reference answers; and precise evaluation through an Agent-as-a-Judge with multiple evaluation metrics. Using FinanceComplexQA, we conduct a comprehensive evaluation of leading RAG systems and agentic reasoning tools for financial document QA. Through identifying and analyzing failure cases, we provide an in-depth study of their capabilities in numerical computation, multi-hop reasoning, content summarization, and industry analysis.

The proliferation of social media platforms has led to an explosion of multimodal data that encapsulates rich emotional content. Effectively integrating heterogeneous modalities to predict sentiment polarity remains a critical challenge. Existing approaches often underexploit sentiment prior knowledge and largely ignore the impact of social links on emotional trends, resulting in suboptimal performance. To address these limitations, we propose a novel multimodal sentiment analysis framework, i.e., Label Semantic Guidance under Social Links (LSGSL). LSGSL enhances sentiment reasoning by jointly modeling visual-textual features and the social relationships between users. Specifically, it encodes social links as a graph structure to facilitate sentiment-aware interactions across modalities, and introduces a novel use of sentiment labels-not merely as classification targets, but as semantic embeddings that guide the fusion and reasoning processes. Furthermore, LSGSL adopts a multi-task learning paradigm that jointly optimizes three objectives: image-text contrastive loss, sentiment-guided semantic similarity loss, and sentiment polarity classification loss. Extensive experiments on three widely-used benchmark datasets demonstrate that LSGSL consistently outperforms state-of-the-art methods, offering new insights into the role of social context and semantic label guidance in multimodal sentiment analysis.

Recent code large language models have achieved remarkable progress on general programming tasks. Nevertheless, their performance degrades significantly in industrial scenarios that require reasoning about hardware semantics, specialized language constructs, and strict resource constraints. To address these challenges, we introduce InCoder-32B (Industrial-Coder-32B), the first 32B-parameter code foundation model unifying code intelligence across chip design, GPU kernel optimization, embedded systems, compiler optimization, and 3D modeling. By adopting an efficient architecture, we train InCoder-32B from scratch with general code pre-training, curated industrial code annealing, mid-training that progressively extends context from 8K to 128K tokens with synthetic industrial reasoning data, and post-training with execution-grounded verification. We conduct extensive evaluation on 14 mainstream general code benchmarks and 9 industrial benchmarks spanning 4 specialized domains. Results show InCoder-32B achieves highly competitive performance on general tasks while establishing strong open-source baselines across industrial domains.

Large language models (LLMs) with in-context learning (ICL) have notably boosted the performance in text-to-SQL, with prior efforts concentrating on exclusive SQL prompts to enhance reasoning ability. However, it is still challenging to further enhance the operational efficiency and inference performance of LLMs. To tackle this challenge, we propose MR-SQL, a multi-level retrieval-based LLM framework consists of three specially designed retrievers. The retrievers collaborate to retrieve valuable information for the target question, which not only reduce schema size and minimize the interference noise, but also enhance the reasoning capability of LLMs through more similar Chains of Thought (CoT). Concretely, Table-Retriever and Column-Retriever retrieve concise tables and columns from original large databases with redundant schema information. Example-Retriever select similar few-shot examples for more targeted CoT. Experiment results indicate that MR-SQL increases the execution accuracy on the BIRD and Spider validation sets by +2.54

Evaluating large language models (LLMs) in safety-critical, physics-governed environments requires more than accuracy-based metrics, because predictions that are numerically close to the ground truth can still violate operational constraints, combine fields in physically inconsistent ways, or fail to produce usable structured outputs. Existing evaluation protocols do not measure these failure modes reliably. We propose FLY-EVAL++, an evidence-driven evaluation protocol that combines deterministic verification of protocol compliance, physical feasibility, and safety constraints with fixed rubric-guided aggregation into interpretable multi-dimensional scores. We instantiate FLY-EVAL++ for Flight Trajectory and Attitude Prediction (FTAP) by extending the PilotBench setting with history-conditioned and multi-step prediction tasks. Across 66 LLMs, safety compliance is the most discriminative dimension of model behavior: models with comparable predictive performance differ by more than 28 points in safety score, and we observe recurrent failures including safety violations under physically plausible predictions and instability in multi-step rollouts. These results show that evaluation in safety-critical domains should measure constraint satisfaction and structured validity explicitly rather than rely on accuracy-centric reporting alone.

Search-and-rescue (SAR) requires embodied agents to explore unfamiliar environments under multimodal uncertainty, perform multi-stage interactions, and retrieve spatial memory over long horizons. Existing benchmarks typically evaluate these capabilities in isolation, leaving unclear how failures compound when they must be composed in realistic workflows. We introduce RescueBench, a photo-realistic diagnostic benchmark that instantiates SAR as a four-stage pipeline: multimodal exploration, target rescue, memory-guided return, and final handoff. By combining sequential task composition with stage-level evaluation, RescueBench enables analysis of how exploration and memory failures propagate through embodied rescue workflows. It contains five progressive difficulty levels that vary in environmental complexity, clue ambiguity, and spatial hierarchy, along with an automatic episode generation and annotation pipeline for scalable evaluation and training. We evaluate seven baselines, an oracle reference, and human players, showing that no baselines complete the full task at the greatest difficulty. Stage-level diagnosis identifies autonomous exploration as the dominant failure mode and spatial memory as a second, independent bottleneck, suggesting that these limitations are not resolved by current topological visual-language navigation or map-based methods. Code is available in https://github.com/wukui-muc/RescueBench

Realizing active visual tracking with a single unified model across diverse robots is challenging, as the physical constraints and motion dynamics vary drastically from one platform to another. Existing approaches typically train separate models for each embodiment, leading to poor scalability and limited generalization. To address this, we propose AdaTracker, an adaptive in-context policy learning framework that robustly tracks targets on diverse robot morphologies. Our key insight is to explicitly model embodiment-specific constraints through an Embodiment Context Encoder, which infers embodiment-specific constraints from history. This contextual representation dynamically modulates a Context-Aware Policy, enabling it to infer optimal control actions for unseen embodiments in a zero-shot manner. To enhance robustness, we introduce two auxiliary objectives to ensure accurate context identification and temporal consistency. Experiments in both simulation and the real world demonstrate that AdaTracker significantly outperforms state-of-the-art methods in cross-embodiment generalization, sample efficiency, and zero-shot adaptation.