We present an empirical study in the geometric task of learning interatomic potentials, which shows equivariance matters even more at larger scales; we show a clear power-law scaling behaviour with respect to data, parameters and compute with ``architecture-dependent exponents''. In particular, we observe that equivariant architectures, which leverage task symmetry, scale better than non-equivariant models. Moreover, among equivariant architectures, higher-order representations translate to better scaling exponents. Our analysis also suggests that for compute-optimal training, the data and model sizes should scale in tandem regardless of the architecture. At a high level, these results suggest that, contrary to common belief, we should not leave it to the model to discover fundamental inductive biases such as symmetry, especially as we scale, because they change the inherent difficulty of the task and its scaling laws.

Patent drafting is complex due to its need for detailed technical descriptions, legal compliance, and visual elements. Although Large Vision Language Models (LVLMs) show promise across various tasks, their application in automating patent writing remains underexplored. In this paper, we present PatentVision, a multimodal framework that integrates textual and visual inputs such as patent claims and drawings to generate complete patent specifications. Built on advanced LVLMs, PatentVision enhances accuracy by combining fine tuned vision language models with domain specific training tailored to patents. Experiments reveal it surpasses text only methods, producing outputs with greater fidelity and alignment with human written standards. Its incorporation of visual data allows it to better represent intricate design features and functional connections, leading to richer and more precise results. This study underscores the value of multimodal techniques in patent automation, providing a scalable tool to reduce manual workloads and improve consistency. PatentVision not only advances patent drafting but also lays the groundwork for broader use of LVLMs in specialized areas, potentially transforming intellectual property management and innovation processes.

Patent drafting presents significant challenges due to its reliance on the extensive experience and specialized expertise of patent attorneys, who must possess both legal acumen and technical understanding of an invention to craft patent applications in a formal legal writing style. This paper presents a demonstration of Patentformer, an AI-powered automated patent drafting platform designed to support patent attorneys by rapidly producing high-quality patent applications adhering to legal writing standards.

Multimodal large language models (MLLMs) often miss small details and spatial relations in cluttered scenes, leading to errors in fine-grained perceptual grounding. We introduce AttWarp, a lightweight method that allocates more resolution to query-relevant content while compressing less informative areas, all while preserving global context. At test time, the approach uses an MLLM's cross-modal attention to perform rectilinear warping of the input image, reallocating spatial resolution toward regions the model deems important, without changing model weights or architecture. This attention-guided warping preserves all original image information but redistributes it non-uniformly, so small objects and subtle relationships become easier for the same model to read while the global layout remains intact. Across five benchmarks (TextVQA, GQA, DocVQA, POPE, MMMU) and four MLLMs (LLaVA, Qwen-VL, InternVL, and InstructBLIP), AttWarp consistently improves accuracy, strengthens compositional reasoning, and reduces hallucinations, outperforming four competitive baselines that manipulate raw images at test time. Together, these results show that attention-guided warping prioritizes information relevant to the query while preserving context, and that the same MLLMs perform better when given such warped inputs.

This research introduces the Judge's Verdict Benchmark, a novel two-step methodology to evaluate Large Language Models (LLMs) as judges for response accuracy evaluation tasks. We assess how well 54 LLMs can replicate human judgment when scoring responses from RAG (Retrieval-Augmented Generation) or Agentic pipelines against ground truth answers. Our methodology progresses from traditional correlation analysis to comprehensive Cohen's Kappa analysis that measures actual agreement patterns. The two-step approach includes: (1) a correlation test that filters judges with strong alignment, followed by (2) a human-likeness test using z-scores to identify two distinct judgment patterns: human-like judgment (|z| < 1) that mimics natural human variation, and super-consistent judgment (z > 1) that exceeds typical human-to-human agreement levels. This methodology reveals that 27 out of 54 tested LLMs achieve Tier 1 performance: 23 models exhibit human-like patterns that preserve the nuances of human judgment, while 4 models demonstrate super-consistent behavior, a pattern that could indicate either enhanced reliability or oversimplification of complex judgments. Testing 43 open-source models (1B-405B parameters) and 11 closed models (GPT, Gemini, Claude variants), we demonstrate that judge excellence is not solely dependent on model size but on specific training strategies. Our key contributions include: (1) establishing that correlation alone is insufficient for judge evaluation, (2) introducing a "Turing Test for judges" based on agreement patterns, and (3) providing a standardized benchmark for classifying LLM judges into distinct performance tiers for different evaluation needs.

Visual retrieval-augmented generation (VRAG) augments vision-language models (VLMs) with external visual knowledge to ground reasoning and reduce hallucinations. Y et current VRAG systems often fail to reliably perceive and integrate evidence across multiple images, leading to weak grounding and erroneous conclusions. In this paper, we propose EVisRAG, an end-to-end framework that learns to reason with evidence-guided multi-image to address this issue. The model first observes retrieved images and records per-image evidence, then derives the final answer from the aggregated evidence. To train EVisRAG effectively, we introduce Reward-Scoped Group Relative Policy Optimization (RS-GRPO), which binds fine-grained rewards to scope-specific tokens to jointly optimize visual perception and reasoning abilities of VLMs. Experimental results on multiple visual question answering benchmarks demonstrate that EVisRAG delivers substantial end-to-end gains over backbone VLM with 27% improvements on average. Further analysis shows that, powered by RS-GRPO, EVisRAG improves answer accuracy by precisely perceiving and localizing question-relevant evidence across multiple images and deriving the final answer from that evidence, much like a real detective. All codes are available at https://github.com/OpenBMB/VisRAG. Retrieval-Augmented Generation (RAG) equips Large Language Models (LLMs) with a knowledge retriever that accesses a curated external knowledge base, supplying task-relevant context at generation time and mitigating hallucinations arising from insufficient parametric knowledge (Lewis et al., 2020; Asai et al., 2024). However, ineffective use of retrieved information limits practical adoption in domain-specific tasks.

Large Language Models (LLMs) are increasingly used to generate textual explanations of process models discovered from event logs. Producing explanations from large behavioral abstractions (e.g., directly-follows graphs or Petri nets) can be computationally expensive. This paper reports an exploratory evaluation of explanation quality under progressive behavioral-input reduction, where models are discovered from progressively smaller prefixes of a fixed log. Our pipeline (i) discovers models at multiple input sizes, (ii) prompts an LLM to generate explanations, and (iii) uses a second LLM to assess completeness, bottleneck identification, and suggested improvements. On synthetic logs, explanation quality is largely preserved under moderate reduction, indicating a practical cost-quality trade-off. The study is exploratory, as the scores are LLM-based (comparative signals rather than ground truth) and the data are synthetic. The results suggest a path toward more computationally efficient, LLM-assisted process analysis in resource-constrained settings.

Large Language Models (LLMs) are increasingly capable of generating complete applications from natural language instructions, creating new opportunities in science and education. In these domains, interactive scientific demonstrations are particularly valuable for explaining concepts, supporting new teaching methods, and presenting research findings. Generating such demonstrations requires models to combine accurate scientific knowledge with the ability to implement interactive front-end code that behaves correctly and responds to user actions. This capability goes beyond the scope of existing benchmarks, which typically evaluate either knowledge question answering without grounding in code or static web code generation without scientific interactivity. To evaluate this integrated ability, we design a hybrid framework that combines programmatic functional testing to rigorously verify interaction logic with visually-grounded qualitative testing to assess rendered outputs against reference snapshots. Building on this framework, we present InteractScience, a benchmark consisting of a substantial set of carefully designed questions across five scientific domains, each paired with unit tests, reference snapshots, and checklists. We evaluate 30 leading open- and closed-source LLMs and report results that highlight ongoing weaknesses in integrating domain knowledge with interactive front-end coding. Our work positions InteractScience as the first benchmark to automatically measure this combined capability with realistic interactive operations, providing a foundation for advancing reliable and educationally useful scientific demonstration code generation. All code and data are publicly available at https://github.com/open-compass/InteractScience.

In this paper, we introduce Narrative Learning, a methodology where models are defined entirely in natural language and iteratively refine their classification criteria using explanatory prompts rather than traditional numerical optimisation. We report on experiments to evaluate the accuracy and potential of this approach using 3 synthetic and 3 natural datasets and compare them against 7 baseline explainable machine learning models. We demonstrate that on 5 out of 6 of these datasets, Narrative Learning became more accurate than the baseline explainable models in 2025 or earlier because of improvements in language models. We also report on trends in the lexicostatistics of these models' outputs as a proxy for the comprehensibility of the explanations.

Automated resume information extraction is critical for scaling talent acquisition, yet its real-world deployment faces three major challenges: the extreme heterogeneity of resume layouts and content, the high cost and latency of large language models (LLMs), and the lack of standardized datasets and evaluation tools. In this work, we present a layout-aware and efficiency-optimized framework for automated extraction and evaluation that addresses all three challenges. Our system combines a fine-tuned layout parser to normalize diverse document formats, an inference-efficient LLM extractor based on parallel prompting and instruction tuning, and a robust two-stage automated evaluation framework supported by new benchmark datasets. Extensive experiments show that our framework significantly outperforms strong baselines in both accuracy and efficiency. In particular, we demonstrate that a fine-tuned compact 0.6B LLM achieves top-tier accuracy while significantly reducing inference latency and computational cost. The system is fully deployed in Alibaba's intelligent HR platform, supporting real-time applications across its business units.