James H. Thorne
YOU?
Author Swipe
View article: Conserving climate‐change refugia: Insights from research and practice
Conserving climate‐change refugia: Insights from research and practice Open
As the impacts of anthropogenic climate change increase, conservation of climate‐change refugia has become a key strategy for effective environmental stewardship. Over the last 5 years, the field of climate‐change refugia conservation has …
View article: When Tom Eats Kimchi: Evaluating Cultural Bias of Multimodal Large Language Models in Cultural Mixture Contexts
When Tom Eats Kimchi: Evaluating Cultural Bias of Multimodal Large Language Models in Cultural Mixture Contexts Open
In a highly globalized world, it is important for multi-modal large language models (MLLMs) to recognize and respond correctly to mixed-cultural inputs. For example, a model should correctly identify kimchi (Korean food) in an image both w…
View article: Phenological Shifts Since 1830 in 29 Native Plant Species of California and Their Responses to Historical Climate Change
Phenological Shifts Since 1830 in 29 Native Plant Species of California and Their Responses to Historical Climate Change Open
Climate change is affecting Mediterranean climate regions, such as California. Retrospective phenological studies are a useful tool to track biological response to these impacts through the use of herbarium-preserved specimens. We used dat…
View article: How Blind and Low-Vision Individuals Prefer Large Vision-Language Model-Generated Scene Descriptions
How Blind and Low-Vision Individuals Prefer Large Vision-Language Model-Generated Scene Descriptions Open
For individuals with blindness or low vision (BLV), navigating complex environments can pose serious risks. Large Vision-Language Models (LVLMs) show promise for generating scene descriptions, but their effectiveness for BLV users remains …
View article: Diffusion Models Through a Global Lens: Are They Culturally Inclusive?
Diffusion Models Through a Global Lens: Are They Culturally Inclusive? Open
Text-to-image diffusion models have recently enabled the creation of visually compelling, detailed images from textual prompts. However, their ability to accurately represent various cultural nuances remains an open question. In our work, …
View article: Parallel Key-Value Cache Fusion for Position Invariant RAG
Parallel Key-Value Cache Fusion for Position Invariant RAG Open
Recent advancements in Large Language Models (LLMs) underscore the necessity of Retrieval Augmented Generation (RAG) to leverage external information. However, LLMs are sensitive to the position of relevant information within contexts and …
View article: Learning to Insert [PAUSE] Tokens for Better Reasoning
Learning to Insert [PAUSE] Tokens for Better Reasoning Open
To enhance reasoning capabilities, previous works have explored incorporating special-purpose tokens into the training process. These strategies strengthen the learning mechanism of transformer-based large language models (LLMs). Building …
View article: I0T: Embedding Standardization Method Towards Zero Modality Gap
I0T: Embedding Standardization Method Towards Zero Modality Gap Open
Contrastive Language-Image Pretraining (CLIP) enables zero-shot inference in downstream tasks such as image-text retrieval and classification. However, recent works extending CLIP suffer from the issue of modality gap, which arises when th…
View article: Context Filtering with Reward Modeling in Question Answering
Context Filtering with Reward Modeling in Question Answering Open
Question Answering (QA) in NLP is the task of finding answers to a query within a relevant context retrieved by a retrieval system. Yet, the mix of relevant and irrelevant information in these contexts can hinder performance enhancements i…
View article: Characterizing Soil and Bedrock Water Use of Native California Vegetation
Characterizing Soil and Bedrock Water Use of Native California Vegetation Open
The effective characterization of landscape water balance components—evapotranspiration, runoff, recharge, and soil storage—is critical for understanding the integrated effects of the water balance on vegetation dynamics, water availabilit…
View article: Fine-scale surficial soil moisture mapping using UAS-based L-band remote sensing in a mixed oak-grassland landscape
Fine-scale surficial soil moisture mapping using UAS-based L-band remote sensing in a mixed oak-grassland landscape Open
Soil moisture maps provide quantitative information that, along with climate and energy balance, is critical to integrate with hydrologic processes for characterizing landscape conditions. However, soil moisture maps are difficult to produ…
View article: The Automated Verification of Textual Claims (AVeriTeC) Shared Task
The Automated Verification of Textual Claims (AVeriTeC) Shared Task Open
The Automated Verification of Textual Claims (AVeriTeC) shared task asks participants to retrieve evidence and predict veracity for real-world claims checked by fact-checkers. Evidence can be found either via a search engine, or via a know…
View article: Cross-lingual Transfer of Reward Models in Multilingual Alignment
Cross-lingual Transfer of Reward Models in Multilingual Alignment Open
Reinforcement learning with human feedback (RLHF) is shown to largely benefit from precise reward models (RMs). However, recent studies in reward modeling schemes are skewed towards English, limiting the applicability of RLHF in multilingu…
View article: Stable Language Model Pre-training by Reducing Embedding Variability
Stable Language Model Pre-training by Reducing Embedding Variability Open
Stable pre-training is essential for achieving better-performing language models. However, tracking pre-training stability by calculating gradient variance at every step is impractical due to the significant computational costs. We explore…
View article: Will there be water? Climate change, housing needs, and future water demand in California
Will there be water? Climate change, housing needs, and future water demand in California Open
Climate change in California is expected to alter future water availability, impacting water supplies needed to support future housing growth and agriculture demand. In groundwater-dependent regions like California's Central Coast, new lan…
View article: Climate change and California’s terrestrial biodiversity
Climate change and California’s terrestrial biodiversity Open
In this review and synthesis, we argue that California is an important test case for the nation and world because terrestrial biodiversity is very high, present and anticipated threats to biodiversity from climate change and other interact…
View article: Uncertainty in consensus predictions of plant species' vulnerability to climate change
Uncertainty in consensus predictions of plant species' vulnerability to climate change Open
Aim Variation in spatial predictions of species' ranges made by various models has been recognized as a significant source of uncertainty for modelling species distributions. Consensus approaches that combine the results of multiple models…
View article: Margin-aware Preference Optimization for Aligning Diffusion Models without Reference
Margin-aware Preference Optimization for Aligning Diffusion Models without Reference Open
Modern preference alignment methods, such as DPO, rely on divergence regularization to a reference model for training stability-but this creates a fundamental problem we call "reference mismatch." In this paper, we investigate the negative…
View article: SPATIAL PATTERNS OF VEGETATION CHANGE IN A FIRE-SUPPRESSED COASTAL CALIFORNIA LANDSCAPE
SPATIAL PATTERNS OF VEGETATION CHANGE IN A FIRE-SUPPRESSED COASTAL CALIFORNIA LANDSCAPE Open
California's central coast contains high species richness and plant endemism that is threatened by ongoing land use and climate change. Better understanding of regional vegetation dynamics is needed, where its vegetation mosaic and stand s…
View article: ORPO: Monolithic Preference Optimization without Reference Model
ORPO: Monolithic Preference Optimization without Reference Model Open
While recent preference alignment algorithms for language models have demonstrated promising results, supervised fine-tuning (SFT) remains imperative for achieving successful convergence. In this paper, we study the crucial role of SFT wit…
View article: Citizen-science data identifies the daily movement patterns and habitat associations of a nocturnal urban-invading bird species (Corvus frugilegus)
Citizen-science data identifies the daily movement patterns and habitat associations of a nocturnal urban-invading bird species (Corvus frugilegus) Open
Rooks ( Corvus frugilegus ) are an invasive bird species in South Korea that are deemed harmful due to nocturnal urban invasions and agricultural damage. Employing citizen science data, we document the daily movement patterns and habitat a…
View article: Re3val: Reinforced and Reranked Generative Retrieval
Re3val: Reinforced and Reranked Generative Retrieval Open
Generative retrieval models encode pointers to information in a corpus as an index within the model's parameters. These models serve as part of a larger pipeline, where retrieved information conditions generation for knowledge-intensive NL…