Spreadsheets are widely used for business analysis, financial modeling, reporting, and decision-making. However, most existing spreadsheet benchmarks evaluate isolated operations such as single-formula generation or local cell edits, and therefore fail to capture end-to-end workflows in realistic business settings. We introduce SpreadsheetBench 2, a workflow-level benchmark for spreadsheet agents that covers three task categories: generation, debugging, and visualization. The benchmark is constructed from authentic business data, including financial reports and corporate filings, and is annotated and validated by domain experts. The benchmark contains 321 tasks; each instance averages 11.8 worksheets and requires 593.5 cell modifications, reflecting large multi-sheet workbooks with cross-sheet dependencies. We evaluate eight frontier large language models under a unified multi-turn agent scaffold, and additionally include several LLM-based spreadsheet products as complementary baselines. Results show that current systems remain far from reliable on real-world workflows: the best model achieves 34.89% overall task accuracy, and debugging accuracy is as low as 12.00%. Trajectory analysis and a failure taxonomy further indicate that insufficient spreadsheet inspection and incorrect target-cell selection are the dominant bottlenecks. Together, these findings position SpreadsheetBench 2 as a challenging testbed for advancing reliable spreadsheet automation. Project page: https://spreadsheetbench.github.io/

Text-to-SQL aims to translate natural language questions into executable SQL queries over structured databases. Existing benchmarks mainly focus on closed-domain settings with predefined database schemas and well-specified questions, but they fall short in addressing the challenges of open-domain scenarios, such as ambiguous questions, unspecified databases, and cross-database querying. To bridge this gap, we introduce TACO, a benchmark for open-domain Text-to-SQL with Ambiguous and Cross-database queries. TACO consists of 1,500 real-world Text-to-SQL examples based on a smart city data service and 13,000 high-quality synthetic examples generated based on large-scale open data portals, covering diverse domains such as transportation, healthcare, and finance. To construct the synthetic examples, we develop an effective data synthesis pipeline that preserves the complexity of real-world queries. To demonstrate the utility of TACO, we introduce a baseline TACO-SQL composed of question rewriting, table linking, and query planning, to illustrate the challenges posed by TACO and to better understand the limitations of existing Text-to-SQL approaches. Extensive experiments on TACO using a variety of recent Text-to-SQL approaches show that, while TACO-SQL achieves the best results, a significant gap still remains between the existing approaches and human-written SQL. These findings highlight the difficulty of open-domain Text-to-SQL and position TACO as a valuable benchmark to drive future research.

Vision-Language-Action (VLA) models have recently demonstrated strong performance across embodied tasks. Modern VLAs commonly employ diffusion action experts to efficiently generate high-precision continuous action chunks, while auto-regressive generation can be slower and less accurate at low-level control. Yet auto-regressive paradigms still provide complementary priors that can improve robustness and generalization in out-of-distribution environments. To leverage both paradigms, we propose Action-Draft-and-Verify (ADV): diffusion action expert drafts multiple candidate action chunks, and the VLM selects one by scoring all candidates in a single forward pass with a perplexity-style metric. Under matched backbones, training data, and action-chunk length, ADV improves success rate by +4.3 points in simulation and +19.7 points in real-world over diffusion-based baseline, with a single-pass VLM reranking overhead.

Existing Low-Rank Adaptation (LoRA) methods face challenges on sparse Large Language Models (LLMs) due to the inability to maintain sparsity. Recent works introduce methods that maintain sparsity by augmenting LoRA techniques with additional masking mechanisms. Despite these successes, such approaches suffer from an increased memory and computation overhead, which affects the efficiency of LoRA methods. In response to this limitation, we introduce Low Rank adaptation method for Sparse LLM (LoRS), an innovative method designed to achieve both memory and computation efficiency when fine-tuning sparse LLMs. To mitigate the substantial memory and computation demands associated with preserving sparsity, our approach incorporates strategies of weight recomputing and computational graph rearrangement. In addition, we also improve the effectiveness of LoRS through better adapter initialization. These innovations lead to a notable reduction in memory and computation consumption during the fine-tuning phase, while achieving performance levels that outperform existing LoRA approaches.

Long-horizon robotic manipulation requires dense feedback that reflects how a task advances through its procedural stages, not merely whether the final outcome is successful. Existing reward models often rely on trajectory-level success labels or time-based interpolation, which can conflate elapsed time with true task progress and therefore fail to capture unfinished steps, stagnation, and failure states. We present ProcVLM, a progress-aware vision-language model that learns procedure-grounded progress as a dense reward signal for manipulation. Rather than deriving progress from terminal outcomes or temporal proxies, ProcVLM grounds progress estimation in procedural structure and intra-stage visual change, and further adopts a reasoning-before-estimation paradigm that infers the remaining atomic actions before estimating task progress. Specifically, we construct this supervision by synthesizing frame-level subtask-semantic annotations, assigning progress budgets according to subtask structure, and distributing each budget based on intra-subtask visual change. To train ProcVLM at scale, we build a standardized procedural supervision synthesis pipeline and construct ProcCorpus-60M from 30 embodied datasets with 60M annotated frames, from which we derive ProcVQA for procedure-aware pretraining, with progress estimation as the central task alongside action segmentation and future planning. Experiments on ProcVQA and reward-model benchmarks show that ProcVLM improves embodied procedural reasoning and yields more discriminative trajectory-internal progress estimates than representative baselines, supporting its use as a dense reward model for downstream reward-guided policy optimization. Project page: https://procvlm.github.io/

In this work, we conduct a systematic analysis of Native Sparse Attention (NSA) and propose targeted improvements that enhance long-context modeling. A key insight is that alternating between local (sliding-window) and global (compression, selective) attention across layers, rather than using fixed patterns, enables more effective propagation of long-range dependencies and substantially boosts performance on long-sequence tasks. Meanwhile, we further refine NSA's branches with Latent Attention that the sliding-window branch is enhanced with Multi-head Latent Attention (MLA) while compression and selective branches adopt Group-head Latent Attention (GLA). These changes reduce KV-cache memory by 50\% versus NSA while improving the model's common-sense reasoning and long-text understanding capabilities. Experiments on models from 340M to 1.3B parameters (trained on 15B and 100B tokens) show our method matches or exceeds full attention and native sparse attention in both common-sense reasoning and long-context understanding tasks.

The rapid advancement of large language models has transformed survey writing from a months-long manual effort into an automated process. As generation scales, reliable evaluation becomes the bottleneck, and LLMs are increasingly used as survey evaluators. However, existing approaches largely rely on off-the-shelf LLM-as-a-judge methods without systematic alignment to human reviewers, and there remains a lack of systematic frameworks for quantifying alignment with human reviewers. To address this gap, we propose SurveyReview, a reviewer-aligned, multi-dimensional benchmark and dataset for survey evaluation. We collect and annotate 675 survey papers with 1,630 review reports. We structure authentic peer-review reports by converting free-form comments into four-dimensional scores (Readability, Criticalness, Comprehensiveness, Structure) paired with supporting rationales. We further release standardized train/test splits and an evaluation protocol to measure alignment between automatic evaluators and human reviewers. To validate the benchmark, we develop SurveyAlign, a strong baseline evaluator by fine-tuning Qwen3-32B with LoRA on our annotated data, augmented with external knowledge for knowledge-intensive dimensions. On the test set, SurveyAlign substantially improves reviewer alignment over prompt-based judging with GPT-5.2, reducing average MSE from 2.28 to 1.38 and MAE from 1.15 to 0.69 across all four dimensions. Our contributions are twofold: (1) we establish the first multi-dimensional, reviewer-aligned dataset with a reproducible evaluation framework for survey reviewing; (2) we develop a strong baseline evaluator that substantially improves alignment with human reviewers, providing a competitive reference for future research. Our code and data are available at https://surveyreview.github.io

Index recommendation is crucial for optimizing database performance. However, existing heuristic- and learningbased methods often rely on inefficient exhaustive search and estimated costs, leading to low efficiency (due to the vast search space) and unsatisfactory actual latency (due to inaccurate estimations). Inspired by the refinement strategies of experienced DBAs-who efficiently identify and iteratively refine indexes with database feedback-we present LLMIA, an out-of-thebox, tuning-free index advisor leveraging large language models (LLMs) through in-context learning for index recommendation. LLMIA injects database expertise into the LLM using a highquality demonstration pool and comprehensive workload feature extraction, while iteratively incorporating database feedback to guide the index refinement. This design enables LLMIA to emulate the decision-making process of expert DBAs: efficiently recommending and refining indexes for various workloads within just a few interactions with the DBMS. We validate LLMIA with extensive experiments on five standard OLAP benchmarks (TPC-H with different scales, JOB, TPC-DS, SSB), where it consistently outperforms or matches 12 baselines by producing superior index recommendations with minimal database interactions. Additionally, LLMIA demonstrates robust generalization on two real-world commercial workloads, delivering high-quality recommendations without the need for additional adaptation or retraining, highlighting its out-of-the-box capability.

Robust robot control benefits from explicitly modeling state transitions, but video-generation world action models (WAMs) introduce substantial deployment cost. Existing latent WAMs avoid explicit future generation, but often compress predictive representations or separate predictive modeling from the representations used for action generation. We introduce JEPA-WAM, a latent WAM built in a pretrained V-JEPA space, which couples latent transition prediction with continuous action generation through a shared predictor. JEPA-WAM predicts a spatially structured joint current-future target that captures task-shared visual temporal structure between current and future observations, while preserving dense patch-level correspondence. Through the shared predictor, transition supervision directly shapes the backbone, from which dedicated representations are extracted for action prediction. The same design can also be instantiated in pretrained VLA policies while preserving their original perception and action pathways. On LIBERO-Plus, JEPA-WAM achieves 79.2

Large language models (LLMs) have demonstrated strong coding capabilities but still struggle to solve competitive programming problems correctly in a single attempt. Execution-based re-ranking offers a promising test-time scaling strategy, yet existing methods are constrained by either difficult test case generation or inefficient random input sampling. To address this limitation, we propose Agentic Verifier, an execution-based agent that actively reasons about program behaviors and searches for highly discriminative test inputs that expose behavioral discrepancies among candidate solutions. Through multi-turn interaction with code execution environments, the verifier iteratively refines the candidate input generator and produces targeted counterexamples rather than blindly sampling inputs. We train the verifier to acquire this discriminative input generation capability via a scalable pipeline combining large-scale data synthesis, rejection fine-tuning, and agentic reinforcement learning. Extensive experiments across five competitive programming benchmarks demonstrate consistent improvements over strong execution-based baselines, achieving up to +10-15

Latent actions serve as an intermediate representation that enables consistent modeling of vision-language-action (VLA) models across heterogeneous datasets. However, approaches to supervising VLAs with latent actions are fragmented and lack a systematic comparison. This work structures the study of latent action supervision from two perspectives: (i) regularizing the trajectory via image-based latent actions, and (ii) unifying the target space with action-based latent actions. Under a unified VLA baseline, we instantiate and compare four representative integration strategies. Our results reveal a formulation-task correspondence: image-based latent actions benefit long-horizon reasoning, whereas action-based latent actions excel at complex motor coordination. Furthermore, we find that directly supervising the VLM with discrete latent action tokens yields the most effective performance. Finally, our experiments offer initial insights into the benefits of latent action supervision in mixed-data, suggesting a promising direction for VLA training.

Reinforcement Learning (RL) has become essential for eliciting complex reasoning capabilities in Large Language Models (LLMs). However, the substantial memory overhead of storing Key-Value (KV) caches during long-horizon rollouts acts as a critical bottleneck, often prohibiting efficient training on limited hardware. While existing KV compression techniques offer a remedy for inference, directly applying them to RL training induces a severe policy mismatch, leading to catastrophic performance collapse. To address this, we introduce Sparse-RL, which empowers stable RL training under sparse rollouts. We show that instability arises from a fundamental policy mismatch among the dense old policy, the sparse sampler policy, and the learner policy. To mitigate this issue, Sparse-RL incorporates Sparsity-Aware Rejection Sampling and Importance-based Reweighting to correct the off-policy bias introduced by compression-induced information loss. Experimental results show that Sparse-RL reduces rollout overhead compared to dense baselines while preserving the performance. Furthermore, Sparse-RL inherently implements sparsity-aware training, significantly enhancing model robustness during sparse inference deployment.

Large Language Models (LLMs) have been increasingly adopted in Text-to-SQL systems, yet SQL errors remain a major obstacle in real-world Text-to-SQL inference pipelines. Existing SQL correction approaches either rely on large-scale, high-quality training data with substantial overhead, or adopt single-path agentic workflows that are brittle to early mistakes and prone to error propagation. To develop a practical SQL correctness system for industrial scenarios, we present a training-free framework that formulates SQL correction as a plan-guided, tree-structured debugging process. By maintaining multiple correction strategies and enabling backtracking, the framework mitigates error accumulation during iterative refinement. We further integrate execution-based verification and clause-level diagnostic tools to support strategy pruning and precise error localization. We evaluate the system on the BIRD-Critic benchmark and observe consistent accuracy gains over strong LLM backbones and representative agent-based baselines, achieving a 9.42

In Text-to-SQL tasks, existing LLM-based methods often include extensive database schemas in prompts, leading to long context lengths and increased prefilling latency. While user queries typically focus on recurrent table sets-offering an opportunity for KV cache sharing across queries-current inference engines, such as SGLang and vLLM, generate redundant prefix cache copies when processing user queries with varying table orders. To address this inefficiency, we propose precomputing table representations as KV caches offline and querying the required ones online. A key aspect of our approach is the computation of table caches while preserving primary foreign key relationships between tables. Additionally, we construct a Table Trie structure to facilitate efficient KV cache lookups during inference. To enhance cache performance, we introduce a cache management system with a query reranking strategy to improve cache hit rates and a computation loading pipeline for parallelizing model inference and cache loading. Experimental results show that our proposed TableCache achieves up to a 3.62x speedup in Time to First Token (TTFT) with negligible performance degradation.

Cross-embodiment transfer in vision-language-action (VLA) models remains challenging because low-level state and action spaces differ fundamentally across robot platforms. We observe that the high-level cognitive process underlying manipulation, including scene perception, object identification, task planning, and sub-task decomposition, is largely shared across embodiments. Based on this observation, we present ZR-0, a 2.6 billion parameter end-to-end VLA model that uses dense Embodied Chain-of-Thought (ECoT) supervision to align cross-embodiment representations within the vision-language model (VLM). ZR-0 adopts a dual-stream architecture: a pre-trained VLM (System 2) generates structured ECoT reasoning during training, while a Diffusion Transformer-based action expert (System 1) produces continuous action chunks via flow matching. The two components are coupled through cross-attention, with an attention mask that restricts the action expert to input prompt features only, enabling ECoT generation to be entirely skipped at inference without any performance loss. ZR-0 is pre-trained on ProcCorpus-60M, a large-scale dataset comprising approximately 60 million frames (approximately 1,000 hours) from over 400K trajectories, with dense ECoT annotations covering 96.8

Cardinality and cost estimation are critical components of query optimization, as they directly influence the construction of efficient physical execution plans. While machine learning-based estimators have achieved notable success, they face several challenges: (1) Training data derived from rigid, template-driven benchmarks exhibits significant distributional divergence from real-world query workloads, a challenge further compounded by the manual template design in exhaustively representing the full spectrum of query patterns. (2) These methods demonstrate limited generalization capabilities, especially in scenarios involving sub-plan estimation or queries that significantly deviate from the training query templates. Furthermore, the inherent inefficiency of operator-level cardinality estimation frequently undermines its applicability for accurate cost estimation. (3) These approaches frequently fail to leverage the rich semantic information and dynamic dependencies between operators. To address these challenges, we propose a novel operator-level cardinality and cost estimator that simultaneously estimates the cardinality and cost of all sub-plans within a query plan. First, we leverage large language models to generate high-quality and diverse SQL queries, which serve as the foundation for pre-training and fine-tuning our model. Second, we introduce a semantic-based operator encoding strategy, augmented with a novel tree-structure-aware neural network, to effectively represent each sub-plan. Third, we propose a specialized loss function tailored for joint cardinality and cost prediction at the operator level, fully utilizing labels from each sub-plan. Extensive experiments on both synthetic and real-world datasets demonstrate that our method consistently outperforms state-of-the-art approaches.

This study explores the associations between artificial intelligence (AI) usage patterns and perceived critical thinking among university students in China, with a focus on three types of AI use: Information Retrieval Use (IRU), Content Generation Use (CGU), and Text Revision Use (TRU). Grounded in Self-Determination Theory (SDT), the research investigates how academic motivation moderates these relationships. Cognitive Load Theory (CLT) provides an interpretive lens for understanding the observed patterns, though load was not directly measured. A cross-sectional survey was conducted with 342 undergraduate students, collecting data on their AI usage behaviors, academic motivation, and perceived critical thinking dispositions. The results reveal that IRU is positively associated with perceived critical thinking, while CGU shows a negative association. TRU showed no significant association. Regarding moderation effects, intrinsic motivation strengthened the positive association between IRU and perceived critical thinking, but, contrary to expectations, also amplified the negative association between CGU and perceived critical thinking. In contrast, extrinsic motivation buffered the negative association of CGU: at high levels of extrinsic motivation, the negative association between CGU and perceived critical thinking was entirely eliminated. These findings indicate that the cognitive correlates of AI use depend on both usage patterns and motivational factors. The study highlights the importance of considering both the type of AI use and students’ motivational orientation when designing educational strategies for AI integration in higher education.

Exploratory Data Analysis (EDA) is an essential yet tedious process for examining a new dataset. To facilitate it, natural language interfaces (NLIs) can help people intuitively explore the dataset via data-oriented questions. However, existing NLIs primarily focus on providing accurate answers to questions, with few offering explanations or presentations of the data analysis pipeline used to uncover the answer. Such presentations are crucial for EDA as they enhance the interpretability and reliability of the answer, while also helping users understand the analysis process and derive insights. To fill this gap, we introduce Urania, a natural language interactive system that is able to visualize the data analysis pipelines used to resolve input questions. It integrates a natural language interface that allows users to explore data via questions, and a novel data-aware question decomposition algorithm that resolves each input question into a data analysis pipeline. This pipeline is visualized in the form of a datamation, with animated presentations of analysis operations and their corresponding data changes. Through two quantitative experiments and expert interviews, we demonstrated that our data-aware question decomposition algorithm outperforms the state-of-the-art technique in terms of execution accuracy, and that Urania can help people explore datasets better. In the end, we discuss the observations from the studies and the potential future works.

Long-context inference in LLMs faces the dual challenges of quadratic attention complexity and prohibitive KV cache memory. While token-level sparse attention offers superior accuracy, its indexing overhead is costly; block-level methods improve efficiency but sacrifice precision. We propose AsyncTLS, a hierarchical sparse attention system that combines coarse-grained block filtering with fine-grained token selection to balance accuracy and efficiency, coupled with an asynchronous offloading engine that overlaps KV cache transfers with computation via temporal locality exploitation. Evaluated on Qwen3 and GLM-4.7-Flash across GQA, and MLA architectures, AsyncTLS achieves accuracy comparable to full attention while delivering 1.2x - 10.0x operator speedups and 1.3x - 4.7x end-to-end throughput improvements on 48k - 96k contexts.