This work presents NEVU, a benchmark for actor-conditioned, event-centric, and direction-aware human value recognition in news-domain texts. NEVU evaluates whether models can infer values from event-structured evidence, attribute them to the correct social actors, and determine their aligned or contradictory direction. Built from 2865 English news articles, NEVU represents news at four semantic levels, from subevents to composite events and full articles. Using a hierarchical taxonomy of 54 fine-grained and 20 coarse-grained values, the benchmark contains 46,589 semantic units, 72,905 annotated unit–actor pairs, and 168,061 directed value instances. NEVU is constructed through a staged LLM-assisted annotation and verification pipeline, with targeted human verification for unresolved cases. Candidate-level acceptance and agreement are further examined through a multi-group assessment. The experiments show that prompting-only models remain limited, whereas LoRA-tuned open-weight models substantially improve performance, with overall Micro-F1 gains of 29.77 and 8.21 percentage points over the strongest prompting-only open-weight and proprietary baselines, respectively. These gains primarily reflect learnability under the NEVU reference-label setting. NEVU provides a structured benchmark for systematic evaluation and supervised adaptation of actor-conditioned human value recognition in English news.
Landslides are known as widespread natural hazards. The formation of landslide dams, or landslide deposits, functions as a part of cascading geohazards. Recent studies have revealed the effects of landslide deposits on aquatic geochemistry and ecosystems. However, the role of landslide deposits in shaping subsurface redox conditions, closely linked to biogeochemical element cycles, remains unexplored. Herein, we applied electrical resistivity tomography (ERT) and self-potential (SP) surveys to investigate the subsurface structure, groundwater flow patterns, and redox conditions of a landslide deposit formed by shallow landslides that occurred during the 2018 Hokkaido Eastern Iburi earthquake in Japan (Mw 6.6). The ERT results revealed the soil–bedrock interface as a zone of 20–40 Ω·m with an abrupt spatial change. The soil zone had a maximum thickness of 8 m and a groundwater table at 2 m below ground. The measured SP values ranged from − 28.4 to 15.4 mV. The SP trend surface, which represents the linear trend in the spatial distribution of SP values, showed that SP values increase with decreasing elevation and in the downstream direction, suggesting a gaining-stream condition. Positive SP anomalies were primarily attributed to streaming potential driven by the hydraulic gradients, whereas negative SP anomalies were associated with redox gradients induced by a biogeobattery system with reducing soil zones. These results suggest that landslide deposits host spatially heterogeneous redox environments. Such environments facilitate the mobilization of redox-sensitive metals such as Mn and Fe, thereby posing risks to downstream hydrochemistry and aquatic ecosystems. To our knowledge, this is the first application of geophysical methods to estimate the spatial patterns of redox conditions within a landslide deposit, providing insights into freshwater resource management and ecosystem functioning in landslide-prone catchments. Integrating geophysical, hydrochemical, and microbiological approaches will enable us to reveal the underexplored environmental impacts of landslide deposits.
Many animals and plants establish intimate symbiotic relationships with specific microorganisms acquired from the environment. Given the immense diversity of environmental microbiomes, selecting appropriate partners from such a vast microbial pool poses a critical challenge for host organisms. To meet this challenge, hosts have evolved sophisticated internal partner-choice mechanisms that ensure stable associations with beneficial microbes. However, because these symbionts primarily inhabit external environments, environmental conditions themselves are also expected to influence the establishment of symbiosis. Despite this expectation, the mechanistic role of external environmental filters in shaping the intended symbiosis remains largely unexplored. Focusing on stink bugs, which acquire their symbiotic bacteria from soil each generation, we investigated how soil properties influence the establishment of gut symbiosis in terrestrial insects. Microbiome analyses confirmed that Burkholderia sensu lato overwhelmingly dominates a specific gut organ in six stink bug species from the superfamilies Coreoidea and Lygaeoidea, including serious agricultural pests (relative abundance ranging from 74.5 to 100
To elucidate the mechanism involved in the surface modification of hexagonal boron nitride (h-BN) via plasma treatment, we evaluated the number of reactive sites available for functionalization by quantifying the dangling bonds formed on h-BN particles during plasma treatment in solution. Electron spin resonance detected the boron dangling bonds formed on h-BN particles; these dangling bonds gradually decreased upon exposure to an atmospheric environment for up to 650 days as a result of oxygen termination, such as hydroxyl functionalization. These results indicated that plasma treatment increased the number of dangling bonds, up to 14-fold per unit mass, on h-BN surfaces as reactive sites for functionalization. The surface number density of boron dangling bonds as a function of air-exposure days could be expressed as Langmuir adsorption kinetics; the surface reactive-site density was estimated to be of 1015 m-2 on the plasma-treated h-BN. Polymer composite materials incorporating such functionalized h-BN after plasma treatment and oxygen termination showed markedly improved dispersion and flexibility. Overall, plasma treatment effectively introduced reactive sites even on chemically stable materials, while subsequent oxygen termination further promoted surface functionalization.
The performance of large language models (LLMs) in program synthesis and mathematical reasoning is fundamentally limited by the quality of their pre-training corpora. We introduce two openly licensed pre-training datasets, released under the Llama 3.3 Community License, that significantly enhance LLM performance by systematically rewriting public data. SwallowCode ($\approx$16.1 billion tokens) refines Python snippets from The-Stack-v2 through a novel four-stage pipeline: syntax validation, pylint-based style filtering, and a two-stage LLM rewriting process that enforces style conformity and transforms snippets into self-contained, algorithmically efficient examples. Unlike prior methods that rely on exclusionary filtering or limited transformations, our transform-and-retain approach refines low-quality code, maximizing data utility. SwallowMath ($\approx$2.3 billion tokens) enhances Finemath-4+ by removing boilerplate, restoring context, and reformatting solutions into concise, step-by-step explanations. Within a fixed 50 billion token training budget, continual pre-training of Llama-3.1-8B with SwallowCode boosts pass@1 by +17.0 on HumanEval and +16.1 on HumanEval+ compared to Stack-Edu, surpassing the baseline model's code generation capabilities. Similarly, substituting SwallowMath yields +12.4 accuracy on GSM8K and +7.6 on MATH. Ablation studies confirm that each pipeline stage contributes incrementally, with rewriting yielding the largest gains. By releasing datasets, prompts, checkpoints, and pipeline code, we ensure reproducibility and provide a transferable transform-and-retain methodology that can be adapted to other base models and LLM rewriting setups.
