Tony Chen

Real-world time-series applications increasingly require models that can handle time series forecasting, context-conditioned prediction, and language-based temporal reasoning. Yet current time-series foundation models remain fragmented across these capabilities: numerical specialists often provide the strongest forecasts, while language-based models offer broader contextual understanding and analysis. A central challenge is to unify these heterogeneous capabilities without reducing their individual performance. We introduce OpenTSLM TeeMoE, a generalist time-series language model that can forecast directly from observed time series, reason over textual context and temporal patterns, and synthesize and refine predictions from external numerical forecasting specialists. We independently train three low-rank experts for forecast aggregation, native forecasting, and temporal analysis over a shared backbone. A learned LoRA mixture-of-experts controller then weights their frozen parameter updates for each request. Our proposed model achieves strong performance on widely used benchmarks for time series forecasting, context-conditioned prediction, and language-based temporal reasoning, ranking among the top three on GIFT-Eval by mean MASE rank, Context is Key by RCRPS, and TimeSeriesExam by accuracy.

Temporal Foundation Models (TFMs) aim to generalize across domains, datasets, and tasks. Yet, their development remains constrained by fragmented, task-specific data formats, annotations, and processing pipelines. We introduce TimeNet, an open-source data standard and scalable infrastructure that decouples temporal data from task definitions and represents signals, metadata, annotations, and supervision in a shared, extensible data model. TimeNet supports multimodal signals with regular, irregular, or ordinal time axes and expresses different task families (including classification, forecasting, temporal localization, question answering, generation, and editing) as reusable views over the same recordings. This shared representation enables heterogeneous time-series datasets to be combined for large-scale model training across domains, modalities, and tasks. We demonstrate TimeNet by transcoding datasets with 1.5M task instances spanning diverse domains, modalities, temporal scales, and forms of supervision, while retaining practical I/O performance relative to native formats. TimeNet enables an existing TFM training pipeline to support joint training on a configurable number of heterogeneous datasets through configuration changes alone. We show this capability by training TFM across multiple datasets and tasks, obtaining a 14% F1 score improvement compared with models trained on individual datasets. These results show that TimeNet provides the data and systems foundation needed to move beyond task- and dataset-specific TFMs toward models that can learn jointly across heterogeneous domains, modalities, temporal scales, and forms of supervision from a common data model.

ECG-Language Models (ELMs) extend recent advances in Multimodal Large Language Models (MLLMs) to automated ECG interpretation. However, most existing ELMs inherit Vision-Language Model (VLM) design choices and rely on pretrained ECG encoders, introducing substantial architectural and training complexity. Inspired by encoder-free VLMs, we introduce ELF, a family of three encoder-free ELMs that remain competitive with, and often outperform, prior state-of-the-art ELMs across two datasets despite substantially simpler architectures and training pipelines. All code and data are available at github.com/ELM-Research/ECG-Language-Models.

Interest in generative Electrocardiogram-Language Models (ELMs) is growing, as they can produce textual responses conditioned on ECG signals and textual queries. Unlike traditional classifiers that output label probabilities, ELMs are more versatile, supporting domain-specific tasks (e.g., waveform analysis, diagnosis, prognosis) as well as general tasks (e.g., open-ended questions, dialogue). Retrieval-Augmented Generation (RAG), widely used in Large Language Models (LLMs) to ground LLM outputs in retrieved knowledge, helps reduce hallucinations and improve natural language generation (NLG). However, despite its promise, no open-source implementation or systematic study of RAG pipeline design for ELMs currently exists. To address this gap, we present the first open-source RAG pipeline for ELMs, along with baselines and ablation studies for NLG. Experiments on three public datasets show that ELMs with RAG consistently improves performance over non-RAG baselines and highlights key ELM design considerations.

Building integrated photovoltaic thermal (BIPVT) greenhouse technology exhibits great potential to enhance crop yielding, energy harvesting, and water management for highly efficient farming and enables controlled environment agriculture (CEA) for automated plant cultivation. The proposed smart greenhouse system uses BIPVT, geothermal, image-based sensing, and control technologies for high energy efficiency and high-fidelity plant growth control. A camera-based, computer-controlled system was demonstrated to successfully monitor single and multiple food plants' growth. The growth rate over the life-cycle of the plant was automatically recorded by cameras without human interference. A novel algorithm was developed to extract the plants from the background of the image and then to measure the length from the ground to the top of the longest plant with different shapes or profiles. The growth rate was automatically generated as fundamental data for farming management and control. Because the plant growth is slow and the errors caused by image noise can be comparable with the actual growth when the interval between images is not large enough, the noise of a camera is evaluated by comparing multiple images under the same exposure condition, which is used to provide the error range of the measurements. The plant growth monitoring system will be integrated into an ongoing greenhouse project for smart greenhouse operation.

• I led the creation of the dataset and designed the end-to-end collection and verification pipeline.
• Each session captures mouse/touch trajectories, timing, and puzzle state at physics-tick resolution (240 Hz physics, 1000 Hz raw input), and can be replayed deterministically.
• Three puzzle types: Sheep Herding (draw paths to guide sheep into a pen), Thread the Needle (navigate through gaps without touching walls), and Polygon Stacking (drag and stack shapes on a platform).
• Privacy-preserving by design: coordinates are normalized to a 200×200 grid, render sizes are randomized per session, and no IPs, cookies, or persistent identifiers are stored.
• Using CaptchaSolve30k in a paper? I'd love to hear about it, so please email me.