Large language models (LLMs) are increasingly used to support creative tasks such as research idea generation. While recent work has shown that structured dialogues between LLMs can improve the novelty and feasibility of generated ideas, the optimal design of such interactions remains unclear. In this study, we conduct a comprehensive analysis of multi-agent LLM dialogues for scientific ideation. We compare different configurations of agent roles, number of agents, and dialogue depth to understand how these factors influence the novelty and feasibility of generated ideas. Our experimental setup includes settings where one agent generates ideas and another critiques them, enabling iterative improvement. Our results show that enlarging the agent cohort, deepening the interaction depth, and broadening agent persona heterogeneity each enrich the diversity of generated ideas. Moreover, specifically increasing critic-side diversity within the ideation-critique-revision loop further boosts the feasibility of the final proposals. Our findings offer practical guidelines for building effective multi-agent LLM systems for scientific ideation. Our code is available at https://github.com/g6000/MultiAgent-Research-Ideator.

Automatic n-gram based metrics such as ROUGE are widely used for evaluating generative tasks such as summarization. While these metrics are considered indicative (even if imperfect) of human evaluation for English, their suitability for other languages remains unclear. To address this, we systematically assess evaluation metrics for generation both n-gram-based and neural based to evaluate their effectiveness across languages and tasks. Specifically, we design a large-scale evaluation suite across eight languages from four typological families: agglutinative, isolating, low-fusional, and high-fusional, spanning both low- and high-resource settings, to analyze their correlation with human judgments. Our findings highlight the sensitivity of evaluation metrics to the language type. For example, in fusional languages, n-gram-based metrics show lower correlation with human assessments compared to isolating and agglutinative languages. We also demonstrate that proper tokenization can significantly mitigate this issue for morphologically rich fusional languages, sometimes even reversing negative trends. Additionally, we show that neural-based metrics specifically trained for evaluation, such as COMET, consistently outperform other neural metrics and better correlate with human judgments in low-resource languages. Overall, our analysis highlights the limitations of n-gram metrics for fusional languages and advocates for greater investment in neural-based metrics trained for evaluation tasks.

The Patent-Based Idea Generation task asks systems to turn real patents into product ideas viable within three years. We propose MK2, a prompt-centric pipeline: Gemini 2.5 drafts and iteratively edits a prompt, grafting useful fragments from weaker outputs; GPT-4.1 then uses this prompt to create one idea per patent, and an Elo loop judged by Qwen3-8B selects the best prompt-all without extra training data. Across three domains, two evaluator types, and six criteria, MK2 topped the automatic leaderboard and won 25 of 36 tests. Only the materials-chemistry track lagged, indicating the need for deeper domain grounding; yet, the results show that lightweight prompt engineering has already delivered competitive, commercially relevant ideation from patents.

This paper examines the impact of Generative Artificial Intelligence (GenAI) on scientific practices, conducting a qualitative review of selected literature to explore its applications, benefits, and challenges. The review draws on the OpenAlex publication database, using a Boolean search approach to identify scientific literature related to GenAI (including large language models and ChatGPT). Thirty-nine highly cited papers and commentaries are reviewed and qualitatively coded. Results are categorized by GenAI applications in science, scientific writing, medical practice, and education and training. The analysis finds that while there is a rapid adoption of GenAI in science and science practice, its long-term implications remain unclear, with ongoing uncertainties about its use and governance. The study provides early insights into GenAI's growing role in science and identifies questions for future research in this evolving field.

Multimodal Large Language Models (MLLMs) have shown strong performance in document image tasks, especially Optical Character Recognition (OCR). However, they struggle with Document Image Machine Translation (DIMT), which requires handling both cross-modal and cross-lingual challenges. Previous efforts to enhance DIMT capability through Supervised Fine-Tuning (SFT) on the DIMT dataset often result in the forgetting of the model's existing monolingual abilities, such as OCR. To address these challenges, we introduce a novel fine-tuning paradigm, named Synchronously Self-Reviewing (SSR) its OCR proficiency, inspired by the concept "Bilingual Cognitive Advantage". Specifically, SSR prompts the model to generate OCR text before producing translation text, which allows the model to leverage its strong monolingual OCR ability while learning to translate text across languages. Comprehensive experiments demonstrate the proposed SSR learning helps mitigate catastrophic forgetting, improving the generalization ability of MLLMs on both OCR and DIMT tasks.

Recent advancements in Multimodal Large Language Models (MLLMs), particularly through Reinforcement Learning with Verifiable Rewards (RLVR), have significantly enhanced their reasoning abilities. However, a critical gap persists: these models struggle with dynamic spatial interactions, a capability essential for real-world applications. To bridge this gap, we introduce M2-Reasoning-7B, a model designed to excel in both general and spatial reasoning. Our approach integrates two key innovations: (1) a novel data pipeline that generates 294.2K high-quality data samples (168K for cold-start fine-tuning and 126.2K for RLVR), which feature logically coherent reasoning trajectories and have undergone comprehensive assessment; and (2) a dynamic multi-task training strategy with step-wise optimization to mitigate conflicts between data, and task-specific rewards for delivering tailored incentive signals. This combination of curated data and advanced training allows M2-Reasoning-7B to set a new state-of-the-art (SOTA) across 8 benchmarks, showcasing superior performance in both general and spatial reasoning domains.

Watermarking technology has gained significant attention due to the increasing importance of intellectual property (IP) rights, particularly with the growing deployment of large language models (LLMs) on billions resource-constrained edge devices. To counter the potential threats of IP theft by malicious users, this paper introduces a robust watermarking scheme without retraining or fine-tuning for transformer models. The scheme generates a unique key for each user and derives a stable watermark value by solving linear constraints constructed from model invariants. Moreover, this technology utilizes noise mechanism to hide watermark locations in multi-user scenarios against collusion attack. This paper evaluates the approach on three popular models (Llama3, Phi3, Gemma), and the experimental results confirm the strong robustness across a range of attack methods (fine-tuning, pruning, quantization, permutation, scaling, reversible matrix and collusion attacks).

We introduce a lightweight yet highly effective safety guardrail framework for language models, demonstrating that small-scale language models can achieve, and even surpass, the performance of larger counterparts in content moderation tasks. This is accomplished through high-fidelity synthetic data generation and adversarial training. The synthetic data generation process begins with human-curated seed data, which undergoes query augmentation and paraphrasing to create diverse and contextually rich examples. This augmented data is then subjected to multiple rounds of curation, ensuring high fidelity and relevance. Inspired by recent advances in the Generative Adversarial Network (GAN) architecture, our adversarial training employs reinforcement learning to guide a generator that produces challenging synthetic examples. These examples are used to fine-tune the safety classifier, enhancing its ability to detect and mitigate harmful content. Additionally, we incorporate strategies from recent research on efficient LLM training, leveraging the capabilities of smaller models to improve the performance of larger generative models. With iterative adversarial training and the generation of diverse, high-quality synthetic data, our framework enables small language models (SLMs) to serve as robust safety guardrails. This approach not only reduces computational overhead but also enhances resilience against adversarial attacks, offering a scalable and efficient solution for content moderation in AI systems.

Enhancing the mathematical reasoning of Large Language Models (LLMs) is a pivotal challenge in advancing AI capabilities. While Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) are the dominant training paradigms, a systematic methodology for combining them to maximize both accuracy and efficiency remains largely unexplored. This paper introduces a practical and effective training recipe that strategically integrates extended SFT with RL from online inference (GRPO). We posit that these methods play complementary, not competing, roles: a prolonged SFT phase first pushes the model's accuracy to its limits, after which a GRPO phase dramatically improves token efficiency while preserving this peak performance. Our experiments reveal that extending SFT for as many as 10 epochs is crucial for performance breakthroughs, and that the primary role of GRPO in this framework is to optimize solution length. The efficacy of our recipe is rigorously validated through top-tier performance on challenging benchmarks, including a high rank among over 2,200 teams in the strictly leak-free AI Mathematical Olympiad (AIMO). This work provides the community with a battle-tested blueprint for developing state-of-the-art mathematical reasoners that are both exceptionally accurate and practically efficient. To ensure full reproducibility and empower future research, we will open-source our entire framework, including all code, model checkpoints, and training configurations at https://github.com/analokmaus/kaggle-aimo2-fast-math-r1.

Accurate identification of fungi species presents a unique challenge in computer vision due to fine-grained inter-species variation and high intra-species variation. This paper presents our approach for the FungiCLEF 2025 competition, which focuses on few-shot fine-grained visual categorization (FGVC) using the FungiTastic Few-Shot dataset. Our team (DS@GT) experimented with multiple vision transformer models, data augmentation, weighted sampling, and incorporating textual information. We also explored generative AI models for zero-shot classification using structured prompting but found them to significantly underperform relative to vision-based models. Our final model outperformed both competition baselines and highlighted the effectiveness of domain specific pretraining and balanced sampling strategies. Our approach ranked 35/74 on the private test set in post-completion evaluation, this suggests additional work can be done on metadata selection and domain-adapted multi-modal learning. Our code is available at https://github.com/dsgt-arc/fungiclef-2025.