We introduce Creativity Benchmark, an evaluation framework for large language models (LLMs) in marketing creativity. The benchmark covers 100 brands (12 categories) and three prompt types (Insights, Ideas, Wild Ideas). Human pairwise preferences from 678 practising creatives over 11,012 anonymised comparisons, analysed with Bradley-Terry models, show tightly clustered performance with no model dominating across brands or prompt types: the top-bottom spread is $Δθ\approx 0.45$, which implies a head-to-head win probability of $0.61$; the highest-rated model beats the lowest only about $61\%$ of the time. We also analyse model diversity using cosine distances to capture intra- and inter-model variation and sensitivity to prompt reframing. Comparing three LLM-as-judge setups with human rankings reveals weak, inconsistent correlations and judge-specific biases, underscoring that automated judges cannot substitute for human evaluation. Conventional creativity tests also transfer only partially to brand-constrained tasks. Overall, the results highlight the need for expert human evaluation and diversity-aware workflows.

--The advancement of robotics and autonomous navigation systems hinges on the ability to accurately predict terrain traversability. Traditional methods for generating datasets to train these prediction models often involve putting robots into potentially hazardous environments, posing risks to equipment and safety. T o solve this problem, we present ZeST, a novel approach leveraging visual reasoning capabilities of Large Language Models (LLMs) to create a traversability map in real-time without exposing robots to danger . Our approach not only performs zero-shot traversability and mitigates the risks associated with real-world data collection but also accelerates the development of advanced navigation systems, offering a cost-effective and scalable solution. T o support our findings, we present navigation results, in both controlled indoor and unstructured outdoor environments. As shown in the experiments, our method provides safer navigation when compared to other state-of-the-art methods, constantly reaching the final goal. The development of autonomous navigation systems is a cornerstone of robotics, with terrain traversability prediction being a critical component [1], [2], [3], [4], [5]. Traversability prediction refers to the ability of a robot to assess whether a given terrain is passable or poses risks to its operation.

Vulnerability databases, such as the National Vulnerability Database (NVD), offer detailed descriptions of Common Vulnerabilities and Exposures (CVEs), but often lack information on their real-world impact, such as the tactics, techniques, and procedures (TTPs) that adversaries may use to exploit the vulnerability. However, manually linking CVEs to their corresponding TTPs is a challenging and time-consuming task, and the high volume of new vulnerabilities published annually makes automated support desirable. This paper introduces TRIAGE, a two-pronged automated approach that uses Large Language Models (LLMs) to map CVEs to relevant techniques from the ATT&CK knowledge base. We first prompt an LLM with instructions based on MITRE's CVE Mapping Methodology to predict an initial list of techniques. This list is then combined with the results from a second LLM-based module that uses in-context learning to map a CVE to relevant techniques. This hybrid approach strategically combines rule-based reasoning with data-driven inference. Our evaluation reveals that in-context learning outperforms the individual mapping methods, and the hybrid approach improves recall of exploitation techniques. We also find that GPT-4o-mini performs better than Llama3.3-70B on this task. Overall, our results show that LLMs can be used to automatically predict the impact of cybersecurity vulnerabilities and TRIAGE makes the process of mapping CVEs to ATT&CK more efficient. A replication package is available for download from https://doi.org/10.5281/zenodo.17341503. Keywords: vulnerability impact, CVE, ATT&CK techniques, large language models, automated mapping.

Sparse Autoencoders (SAEs) can extract interpretable features from large language models (LLMs) without supervision. However, their effectiveness in downstream steering tasks is limited by the requirement for contrastive datasets or large activation storage. To address these limitations, we propose CorrSteer, which selects features by correlating sample correctness with SAE activations from generated tokens at inference time. This approach uses only inference-time activations to extract more relevant features, thereby reducing spurious correlations. It also obtains steering coefficients from average activations, automating the entire pipeline. Our method shows improved task performance on QA, bias mitigation, jailbreaking prevention, and reasoning benchmarks on Gemma-2 2B and LLaMA-3.1 8B, notably achieving a +3.3% improvement in MMLU performance with 4000 samples and a +27.2% improvement in HarmBench with only 108 samples. Selected features demonstrate semantically meaningful patterns aligned with each task's requirements, revealing the underlying capabilities that drive performance. Our work establishes correlation-based selection as an effective and scalable approach for automated SAE steering across language model applications.

The emergence of audio-visual foundation models underscores the importance of reliably assessing their multi-modal understanding. The VGGSound dataset is commonly used as a benchmark for evaluation audio-visual classification. However, our analysis identifies several limitations of VGGSound, including incomplete labelling, partially overlapping classes, and misaligned modalities. These lead to distorted evaluations of auditory and visual capabilities. To address these limitations, we introduce VGGSounder, a comprehensively re-annotated, multi-label test set that extends VGGSound and is specifically designed to evaluate audio-visual foundation models. VGGSounder features detailed modality annotations, enabling precise analyses of modality-specific performance. Furthermore, we reveal model limitations by analysing performance degradation when adding another input modality with our new modality confusion metric.

Recent studies show near-perfect language-based predictions of depression scores (R2 = .70), but these "Mirror" models rely on language responses directly from depression assessments to predict depression assessment scores. These methods suffer from criterion contamination that inflate prediction estimates. We compare "Mirror" models to "Non-Mirror" models, which use other external language to predict depression scores. 110 participants completed both structured diagnostic (Mirror condition) and life history (Non-Mirror condition) interviews. LLMs were prompted to predict diagnostic depression scores. As expected, Mirror models were near-perfect. However, Non-Mirror models also displayed prediction sizes considered large in psychology. Further, both Mirror and Non-Mirror predictions correlated with other questionnaire-based depression symptoms at similar sizes, suggesting bias in Mirror models. Topic modeling revealed different theme structures across model types. As language models for depression continue to evolve, incorporating Non-Mirror approaches may support more valid and clinically useful language-based AI applications in psychological assessment.

Large language models (LLMs) have gained significant attention in chemistry. However, most existing datasets center on molecular-level property prediction and overlook the role of fine-grained functional group (FG) information. Incorporating FG-level data can provide valuable prior knowledge that links molecular structures with textual descriptions, which can be used to build more interpretable, structure-aware LLMs for reasoning on molecule-related tasks. Moreover, LLMs can learn from such fine-grained information to uncover hidden relationships between specific functional groups and molecular properties, thereby advancing molecular design and drug discovery. Here, we introduce FGBench, a dataset comprising 625K molecular property reasoning problems with functional group information. Functional groups are precisely annotated and localized within the molecule, which ensures the dataset's interoperability thereby facilitating further multimodal applications. FGBench includes both regression and classification tasks on 245 different functional groups across three categories for molecular property reasoning: (1) single functional group impacts, (2) multiple functional group interactions, and (3) direct molecular comparisons. In the benchmark of state-of-the-art LLMs on 7K curated data, the results indicate that current LLMs struggle with FG-level property reasoning, highlighting the need to enhance reasoning capabilities in LLMs for chemistry tasks. We anticipate that the methodology employed in FGBench to construct datasets with functional group-level information will serve as a foundational framework for generating new question-answer pairs, enabling LLMs to better understand fine-grained molecular structure-property relationships. The dataset and evaluation code are available at https://github.com/xuanliugit/FGBench.

Graph Neural Networks (GNNs) have achieved remarkable success in graph-based learning by propagating information among neighbor nodes via predefined aggregation mechanisms. However, such fixed schemes often suffer from two key limitations. First, they cannot handle the imbalance in node informativeness -- some nodes are rich in information, while others remain sparse. Second, predefined message passing primarily leverages local structural similarity while ignoring global semantic relationships across the graph, limiting the model's ability to capture distant but relevant information. We propose Retrieval-augmented Graph Agentic Network (ReaGAN), an agent-based framework that empowers each node with autonomous, node-level decision-making. Each node acts as an agent that independently plans its next action based on its internal memory, enabling node-level planning and adaptive message propagation. Additionally, retrieval-augmented generation (RAG) allows nodes to access semantically relevant content and build global relationships in the graph. ReaGAN achieves competitive performance under few-shot in-context settings using a frozen LLM backbone without fine-tuning, showcasing the potential of agentic planning and local-global retrieval in graph learning.

Reinforcement Learning with V erifiable Reward (RL VR) has significantly advanced the complex reasoning abilities of Large Language Models (LLMs). However, it struggles to break through the inherent capability boundaries of the base LLM, due to its essentially on-policy strategy coupled with LLM's immense action space and sparse reward. Critically, RL VR can lead to the capability boundary collapse, narrowing the LLM's problem-solving scope. To address this problem, we propose RL-PLUS, a novel hybrid-policy optimization approach for LLMs that synergizes internal exploitation with external data to achieve stronger reasoning capabilities and surpass the boundaries of base models. RL-PLUS integrates two core components, i.e., Multiple Importance Sampling to address distributional mismatch from external data, and Exploration-Based Advantage Function to guide the model towards high-value, unexplored reasoning paths. We provide both theoretical analysis and extensive experiments to demonstrate the superiority and gen-eralizability of our approach. Compared with existing RL VR methods, RL-PLUS achieves 1) state-of-the-art performance on six math reasoning benchmarks; 2) superior performance on six out-of-distribution reasoning tasks; 3) consistent and significant gains across diverse model families, with average relative improvements up to 69.2%. Moreover, the analysis of Pass@k curves indicates that RL-PLUS effectively resolves the capability boundary collapse problem.

Group Relative Policy Optimization (GRPO) has significantly enhanced the reasoning capability of large language models by optimizing the arithmetic mean of token-level rewards. Unfortunately, GRPO is observed to suffer from unstable policy updates when facing tokens with outlier importance-weighted rewards, which manifest as extreme importance sampling ratios during training. In this study, we propose Geometric-Mean Policy Optimization (GMPO), with the aim to improve the stability of GRPO through suppressing token reward outliers. Instead of optimizing the arithmetic mean, GMPO maximizes the geometric mean of token-level rewards, which is inherently less sensitive to outliers and maintains a more stable range of importance sampling ratio. GMPO is plug-and-play-simply replacing GRPO's arithmetic mean with the geometric mean of token-level rewards, as the latter is inherently less sensitive to outliers. GMPO is theoretically plausible-analysis reveals that both GMPO and GRPO are weighted forms of the policy gradient while the former enjoys more stable weights, which consequently benefits policy optimization and performance. Experiments on multiple mathematical reasoning benchmarks show that GMPO-7B improves the average Pass@1 of GRPO by up to 4.1%, outperforming many state-of-the-art approaches. Code is available at https://github.com/callsys/GMPO.