Samuel Stevens
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Samuel Stevens has a total of 87 co-authors. Here are the top co-authors by number of shared publications:
- Alyson East
Department of Wildlife, Fisheries, and Conservation Biology University of Maine Orono Maine USA
4 shared publications - Charles V. Stewart
Rensselaer Polytechnic Institute
4 shared publications - Sydne Record
Department of Wildlife, Fisheries, and Conservation Biology University of Maine Orono Maine USA
4 shared publications - Tanya Berger-Wolf
Translational Data Analytics Institute The Ohio State University Columbus Ohio USA
3 shared publications Yu Su
The Ohio State University
3 shared publications- Chao Jiang
College of Computing College of Computing Georgia Institute of Technology Georgia Institute of Technology
2 shared publications
Latest Publications
- Open Access
- 11 September 2026
- Published by Springer Nature in Scientific Data
A continental-scale dataset of ground beetles with high-resolution images and validated morphological trait measurements
Despite the ecological significance of invertebrates, global trait databases remain heavily biased toward vertebrates and plants, limiting comprehensive ecological analyses of high-diversity groups like ground beetles. Ground beetles ( Coleoptera: Carabidae ) serve as critical bioindicators of ecosystem health, providing valuable insights into biodiversity shifts driven by environmental changes. While the National Ecological Observatory Network (NEON) maintains an extensive collection of carabid specimens from across the United States, these primarily exist as physical collections, restricting widespread research access and large-scale analysis. To address these gaps, we present a multimodal dataset digitizing over 13,200 NEON carabids from 30 sites spanning the continental US and Hawaii through high-resolution imaging, enabling broader access and computational analysis. The dataset includes human-made measurements of elytra length and width of each specimen image, establishing a foundation for automated trait extraction using AI. High correlations between measurements made by humans from images to those taken by humans on physical specimens ensure reliability for ecological and computational studies. By addressing invertebrate under-representation in trait databases, this work supports AI-driven tools for trait-based research, fostering advancements in biodiversity monitoring and conservation.
- 11 September 2026
- Published by Springer Nature in Scientific Data
Open AccessA continental-scale dataset of ground beetles with high-resolution images and validated morphological trait measurements
Despite the ecological significance of invertebrates, global trait databases remain heavily biased toward vertebrates and plants, limiting comprehensive ecological analyses of high-diversity groups like ground beetles. Ground beetles ( Coleoptera: Carabidae ) serve as critical bioindicators of ecosystem health, providing valuable insights into biodiversity shifts driven by environmental changes. While the National Ecological Observatory Network (NEON) maintains an extensive collection of carabid specimens from across the United States, these primarily exist as physical collections, restricting widespread research access and large-scale analysis. To address these gaps, we present a multimodal dataset digitizing over 13,200 NEON carabids from 30 sites spanning the continental US and Hawaii through high-resolution imaging, enabling broader access and computational analysis. The dataset includes human-made measurements of elytra length and width of each specimen image, establishing a foundation for automated trait extraction using AI. High correlations between measurements made by humans from images to those taken by humans on physical specimens ensure reliability for ecological and computational studies. By addressing invertebrate under-representation in trait databases, this work supports AI-driven tools for trait-based research, fostering advancements in biodiversity monitoring and conservation.
- Open Access
- 31 August 2026
- Published by Pensoft Publishers in ARPHA Preprints
- Vol. 7, e213643
From collection trays to AI-ready data: An operational framework for automated batch entomological specimen processing
Natural history collections house over three billion specimens critical to biodiversity research, yet fewer than 2% of North American arthropod specimens have been imaged, creating a bottleneck for trait-based analyses at broad temporal and spatial scales. Batch photography of multiple specimens improves throughput, but linking detected individuals to database records at scale remains an unsolved operational challenge. The National Ecological Observatory Network (NEON) collects specimens from across the US alongside co-located biotic and abiotic data providing an unprecedented opportunity to explore how digitization efforts can be linked to extended specimen data across space and time. Notably NEON’s curation practice of ordering specimens logically in alignment with metadata enables linking image regions to individual metadata from images containing multiple specimens in a unit tray. We present an accessible framework for high-throughput batch imaging and individual specimen extraction from pinned entomological collections, demonstrated on 65,236 ground beetles (Carabidae) from the NEON Biorepository. The workflow integrates standardized tray photography using consumer equipment, redundant metadata capture by trained technicians, automated validation against the NEON database, and foundation model-based detection and segmentation (Grounding DINO, SAM) to produce individual specimen images with validated metadata linkages. Automated logical queries successfully linked specimens in 78% of trays; the remaining 23% were resolved through structured manual review. The iterative detection pipeline achieved 96% tray-level accuracy across 3,649 trays spanning 518 species. Additionally, this pipeline enables the detection of minor curatorial errors providing an opportunity for imaging and metadata linkage efforts to give back to the collections housing specimens by flagging areas in need of curatorial attention. Complete code, documentation, and standard operating procedures are provided.
- 31 August 2026
- Published by Pensoft Publishers in ARPHA Preprints
- Vol. 7, e213643
Open AccessFrom collection trays to AI-ready data: An operational framework for automated batch entomological specimen processing
Natural history collections house over three billion specimens critical to biodiversity research, yet fewer than 2% of North American arthropod specimens have been imaged, creating a bottleneck for trait-based analyses at broad temporal and spatial scales. Batch photography of multiple specimens improves throughput, but linking detected individuals to database records at scale remains an unsolved operational challenge. The National Ecological Observatory Network (NEON) collects specimens from across the US alongside co-located biotic and abiotic data providing an unprecedented opportunity to explore how digitization efforts can be linked to extended specimen data across space and time. Notably NEON’s curation practice of ordering specimens logically in alignment with metadata enables linking image regions to individual metadata from images containing multiple specimens in a unit tray. We present an accessible framework for high-throughput batch imaging and individual specimen extraction from pinned entomological collections, demonstrated on 65,236 ground beetles (Carabidae) from the NEON Biorepository. The workflow integrates standardized tray photography using consumer equipment, redundant metadata capture by trained technicians, automated validation against the NEON database, and foundation model-based detection and segmentation (Grounding DINO, SAM) to produce individual specimen images with validated metadata linkages. Automated logical queries successfully linked specimens in 78% of trays; the remaining 23% were resolved through structured manual review. The iterative detection pipeline achieved 96% tray-level accuracy across 3,649 trays spanning 518 species. Additionally, this pipeline enables the detection of minor curatorial errors providing an opportunity for imaging and metadata linkage efforts to give back to the collections housing specimens by flagging areas in need of curatorial attention. Complete code, documentation, and standard operating procedures are provided.
- Open Access
Towards Open-Ended Visual Scientific Discovery with Sparse Autoencoders
Scientific archives now contain hundreds of petabytes of data across genomics, ecology, climate, and molecular biology that could reveal undiscovered patterns if systematically analyzed at scale. Large-scale, weakly-supervised datasets in language and vision have driven the development of foundation models whose internal representations encode structure (patterns, co-occurrences and statistical regularities) beyond their training objectives. Most existing methods extract structure only for pre-specified targets; they excel at confirmation but do not support open-ended discovery of unknown patterns. We ask whether sparse autoencoders (SAEs) can enable open-ended feature discovery from foundation model representations. We evaluate this question in controlled rediscovery studies, where the learned SAE features are tested for alignment with semantic concepts on a standard segmentation benchmark and compared against strong label-free alternatives on concept-alignment metrics. Applied to ecological imagery, the same procedure surfaces fine-grained anatomical structure without access to segmentation or part labels, providing a scientific case study with ground-truth validation. While our experiments focus on vision with an ecology case study, the method is domain-agnostic and applicable to models in other sciences (e.g., proteins, genomics, weather). Our results indicate that sparse decomposition provides a practical instrument for exploring what scientific foundation models have learned, an important prerequisite for moving from confirmation to genuine discovery.
Open AccessTowards Open-Ended Visual Scientific Discovery with Sparse Autoencoders
Scientific archives now contain hundreds of petabytes of data across genomics, ecology, climate, and molecular biology that could reveal undiscovered patterns if systematically analyzed at scale. Large-scale, weakly-supervised datasets in language and vision have driven the development of foundation models whose internal representations encode structure (patterns, co-occurrences and statistical regularities) beyond their training objectives. Most existing methods extract structure only for pre-specified targets; they excel at confirmation but do not support open-ended discovery of unknown patterns. We ask whether sparse autoencoders (SAEs) can enable open-ended feature discovery from foundation model representations. We evaluate this question in controlled rediscovery studies, where the learned SAE features are tested for alignment with semantic concepts on a standard segmentation benchmark and compared against strong label-free alternatives on concept-alignment metrics. Applied to ecological imagery, the same procedure surfaces fine-grained anatomical structure without access to segmentation or part labels, providing a scientific case study with ground-truth validation. While our experiments focus on vision with an ecology case study, the method is domain-agnostic and applicable to models in other sciences (e.g., proteins, genomics, weather). Our results indicate that sparse decomposition provides a practical instrument for exploring what scientific foundation models have learned, an important prerequisite for moving from confirmation to genuine discovery.
- Open Access
BioBench: A Blueprint to Move Beyond ImageNet for Scientific ML Benchmarks
ImageNet-1K linear-probe transfer accuracy remains the default proxy for visual representation quality, yet it no longer predicts performance on scientific imagery. Across 46 modern vision model checkpoints, ImageNet top-1 accuracy explains only 34% of variance on ecology tasks and mis-ranks 30% of models above 75% accuracy. We present BioBench, an open ecology vision benchmark that captures what ImageNet misses. BioBench unifies 9 publicly released, application-driven tasks, 4 taxonomic kingdoms, and 6 acquisition modalities (drone RGB, web video, micrographs, in-situ and specimen photos, camera-trap frames), totaling 3.1M images. A single Python API downloads data, fits lightweight classifiers to frozen backbones, and reports class-balanced macro-F1 (plus domain metrics for FishNet and FungiCLEF); ViT-L models evaluate in 6 hours on an A6000 GPU. BioBench provides new signal for computer vision in ecology and a template recipe for building reliable AI-for-science benchmarks in any domain. Code and predictions are available at https://github.com/samuelstevens/biobench and results at https://samuelstevens.me/biobench.
Open AccessBioBench: A Blueprint to Move Beyond ImageNet for Scientific ML Benchmarks
ImageNet-1K linear-probe transfer accuracy remains the default proxy for visual representation quality, yet it no longer predicts performance on scientific imagery. Across 46 modern vision model checkpoints, ImageNet top-1 accuracy explains only 34% of variance on ecology tasks and mis-ranks 30% of models above 75% accuracy. We present BioBench, an open ecology vision benchmark that captures what ImageNet misses. BioBench unifies 9 publicly released, application-driven tasks, 4 taxonomic kingdoms, and 6 acquisition modalities (drone RGB, web video, micrographs, in-situ and specimen photos, camera-trap frames), totaling 3.1M images. A single Python API downloads data, fits lightweight classifiers to frozen backbones, and reports class-balanced macro-F1 (plus domain metrics for FishNet and FungiCLEF); ViT-L models evaluate in 6 hours on an A6000 GPU. BioBench provides new signal for computer vision in ecology and a template recipe for building reliable AI-for-science benchmarks in any domain. Code and predictions are available at https://github.com/samuelstevens/biobench and results at https://samuelstevens.me/biobench.
- Open Access
- 3 September 2025
- Published by Wiley in Methods in Ecology and Evolution
- Vol. 16 (10) , 2260-2275
Optimizing image capture for computer vision‐powered taxonomic identification and trait recognition of biodiversity specimens
Biological collections house millions of specimens with digital images increasingly available through open‐access platforms. However, most imaging protocols were developed for human interpretation without considering automated analysis requirements. As computer vision applications revolutionize taxonomic identification and trait extraction, a critical gap exists between current digitization practices and computational analysis needs. This review provides the first comprehensive practical framework for optimizing biological specimen imaging for computer vision applications. Through interdisciplinary collaboration between taxonomists, collection managers, ecologists and computer scientists, we synthesized evidence‐based recommendations addressing fundamental computer vision concepts and practical imaging considerations. We provide immediately actionable implementation guidance while identifying critical areas requiring community standards development. Our framework encompasses 10 interconnected considerations for optimizing image capture for computer vision‐powered taxonomic identification and trait extraction. We translate these into practical implementation checklists, equipment selection guidelines and a roadmap for community standards development, including filename conventions, pixel density requirements and cross‐institutional protocols. By bridging biological and computational disciplines, this approach unlocks automated analysis potential for millions of existing specimens and guides future digitization efforts towards unprecedented analytical capabilities.
- 3 September 2025
- Published by Wiley in Methods in Ecology and Evolution
- Vol. 16 (10) , 2260-2275
Open AccessOptimizing image capture for computer vision‐powered taxonomic identification and trait recognition of biodiversity specimens
Biological collections house millions of specimens with digital images increasingly available through open‐access platforms. However, most imaging protocols were developed for human interpretation without considering automated analysis requirements. As computer vision applications revolutionize taxonomic identification and trait extraction, a critical gap exists between current digitization practices and computational analysis needs. This review provides the first comprehensive practical framework for optimizing biological specimen imaging for computer vision applications. Through interdisciplinary collaboration between taxonomists, collection managers, ecologists and computer scientists, we synthesized evidence‐based recommendations addressing fundamental computer vision concepts and practical imaging considerations. We provide immediately actionable implementation guidance while identifying critical areas requiring community standards development. Our framework encompasses 10 interconnected considerations for optimizing image capture for computer vision‐powered taxonomic identification and trait extraction. We translate these into practical implementation checklists, equipment selection guidelines and a roadmap for community standards development, including filename conventions, pixel density requirements and cross‐institutional protocols. By bridging biological and computational disciplines, this approach unlocks automated analysis potential for millions of existing specimens and guides future digitization efforts towards unprecedented analytical capabilities.
- Open Access
- 22 July 2025
- Published by MDPI AG in Preprints.org
Fine-Grained Taxonomy with Vision Models: A Benchmark on Long-Tailed and Domain-Adaptive Classification
Ground beetles are a highly sensitive and speciose biological indicator, critical for biodiversity monitoring, yet their taxonomic classification remains underutilized due to the manual effort required for species differentiation based on subtle morphological variations. In this paper, we present a benchmark for fine-grained taxonomic classification, evaluating 12 vision models, across four diverse, long-tailed datasets spanning over 230 genera and 1769 species. These datasets include both controlled laboratory images and challenging field-collected (in-situ) photographs. We investigate two key real-world challenges: sample efficiency and domain adaptation. Our results show that 1) a Vision and Language Transformer with an MLP head achieves best performance, with 97% genus-level and 94% species-level accuracy; 2) efficient subsampling allows train data to be cut in half with minimal performance degradation; 3) model performance significantly drops in domain shift from lab to in-situ settings, highlighting a critical domain gap. Overall, our study lays a foundation for scalable, fine-grained taxonomic classification of beetles and supports broader applications in sample-efficient and domain-adaptive learning for ecological computer vision.
- 22 July 2025
- Published by MDPI AG in Preprints.org
Open AccessFine-Grained Taxonomy with Vision Models: A Benchmark on Long-Tailed and Domain-Adaptive Classification
Ground beetles are a highly sensitive and speciose biological indicator, critical for biodiversity monitoring, yet their taxonomic classification remains underutilized due to the manual effort required for species differentiation based on subtle morphological variations. In this paper, we present a benchmark for fine-grained taxonomic classification, evaluating 12 vision models, across four diverse, long-tailed datasets spanning over 230 genera and 1769 species. These datasets include both controlled laboratory images and challenging field-collected (in-situ) photographs. We investigate two key real-world challenges: sample efficiency and domain adaptation. Our results show that 1) a Vision and Language Transformer with an MLP head achieves best performance, with 97% genus-level and 94% species-level accuracy; 2) efficient subsampling allows train data to be cut in half with minimal performance degradation; 3) model performance significantly drops in domain shift from lab to in-situ settings, highlighting a critical domain gap. Overall, our study lays a foundation for scalable, fine-grained taxonomic classification of beetles and supports broader applications in sample-efficient and domain-adaptive learning for ecological computer vision.
- 21 December 2024
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 84 (21) , 24563-24582
Deep dive into KABR: a dataset for understanding ungulate behavior from in-situ drone video
- 21 December 2024
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 84 (21) , 24563-24582
Deep dive into KABR: a dataset for understanding ungulate behavior from in-situ drone video
- Open Access
Memorization for Good: Encryption with Autoregressive Language Models
Over-parameterized neural language models (LMs) can memorize and recite long sequences of training data. While such memorization is normally associated with undesired properties such as overfitting and information leaking, our work casts memorization as an unexplored capability of LMs. We propose the first symmetric encryption algorithm with autoregressive language models (SELM). We show that autoregressive LMs can encode arbitrary data into a compact real-valued vector (i.e., encryption) and then losslessly decode the vector to the original message (i.e., decryption) via random subspace optimization and greedy decoding. While SELM is not amenable to conventional cryptanalysis, we investigate its security through a novel empirical variant of the classic IND-CPA (indistinguishability under chosen-plaintext attack) game and show promising results on security. Our code and datasets are available at https://github.com/OSU-NLP-Group/SELM.
Open AccessMemorization for Good: Encryption with Autoregressive Language Models
Over-parameterized neural language models (LMs) can memorize and recite long sequences of training data. While such memorization is normally associated with undesired properties such as overfitting and information leaking, our work casts memorization as an unexplored capability of LMs. We propose the first symmetric encryption algorithm with autoregressive language models (SELM). We show that autoregressive LMs can encode arbitrary data into a compact real-valued vector (i.e., encryption) and then losslessly decode the vector to the original message (i.e., decryption) via random subspace optimization and greedy decoding. While SELM is not amenable to conventional cryptanalysis, we investigate its security through a novel empirical variant of the classic IND-CPA (indistinguishability under chosen-plaintext attack) game and show promising results on security. Our code and datasets are available at https://github.com/OSU-NLP-Group/SELM.
- Open Access
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 197-201
Roll Up Your Sleeves: Working with a Collaborative and Engaging Task-Oriented Dialogue System
Lingbo Mo, Shijie Chen, Ziru Chen, Xiang Deng, Ashley Lewis, Sunit Singh, Samuel Stevens, Chang-You Tai, Zhen Wang, Xiang Yue, Tianshu Zhang, Yu Su, Huan Sun. Proceedings of the 24th Meeting of the Special Interest Group on Discourse and Dialogue. 2023.
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 197-201
Open AccessRoll Up Your Sleeves: Working with a Collaborative and Engaging Task-Oriented Dialogue System
Lingbo Mo, Shijie Chen, Ziru Chen, Xiang Deng, Ashley Lewis, Sunit Singh, Samuel Stevens, Chang-You Tai, Zhen Wang, Xiang Yue, Tianshu Zhang, Yu Su, Huan Sun. Proceedings of the 24th Meeting of the Special Interest Group on Discourse and Dialogue. 2023.
- Open Access
arXivEdits: Understanding the Human Revision Process in Scientific Writing
Scientific publications are the primary means to communicate research discoveries, where the writing quality is of crucial importance. However, prior work studying the human editing process in this domain mainly focused on the abstract or introduction sections, resulting in an incomplete picture. In this work, we provide a complete computational framework for studying text revision in scientific writing. We first introduce arXivEdits, a new annotated corpus of 751 full papers from arXiv with gold sentence alignment across their multiple versions of revision, as well as fine-grained span-level edits and their underlying intentions for 1,000 sentence pairs. It supports our data-driven analysis to unveil the common strategies practiced by researchers for revising their papers. To scale up the analysis, we also develop automatic methods to extract revision at document-, sentence-, and word-levels. A neural CRF sentence alignment model trained on our corpus achieves 93.8 F1, enabling the reliable matching of sentences between different versions. We formulate the edit extraction task as a span alignment problem, and our proposed method extracts more fine-grained and explainable edits, compared to the commonly used diff algorithm. An intention classifier trained on our dataset achieves 78.9 F1 on the fine-grained intent classification task. Our data and system are released at tiny.one/arxivedits.
Open AccessarXivEdits: Understanding the Human Revision Process in Scientific Writing
Scientific publications are the primary means to communicate research discoveries, where the writing quality is of crucial importance. However, prior work studying the human editing process in this domain mainly focused on the abstract or introduction sections, resulting in an incomplete picture. In this work, we provide a complete computational framework for studying text revision in scientific writing. We first introduce arXivEdits, a new annotated corpus of 751 full papers from arXiv with gold sentence alignment across their multiple versions of revision, as well as fine-grained span-level edits and their underlying intentions for 1,000 sentence pairs. It supports our data-driven analysis to unveil the common strategies practiced by researchers for revising their papers. To scale up the analysis, we also develop automatic methods to extract revision at document-, sentence-, and word-levels. A neural CRF sentence alignment model trained on our corpus achieves 93.8 F1, enabling the reliable matching of sentences between different versions. We formulate the edit extraction task as a span alignment problem, and our proposed method extracts more fine-grained and explainable edits, compared to the commonly used diff algorithm. An intention classifier trained on our dataset achieves 78.9 F1 on the fine-grained intent classification task. Our data and system are released at tiny.one/arxivedits.