Search for: All records

The field of relation extraction (RE) is experiencing a notable shift towards generative relation extraction (GRE), leveraging the capabilities of large language models (LLMs). However, we discovered that traditional relation extraction (RE) metrics like precision and recall fall short in evaluating GRE methods. This shortfall arises because these metrics rely on exact matching with human-annotated reference relations, while GRE methods often produce diverse and semantically accurate relations that differ from the references. To fill this gap, we introduce GENRES for a multidimensional assessment in terms of the topic similarity, uniqueness, granularity, factualness, and completeness of the GRE results. With GENRES, we empirically identified that (1) precision/recall fails to justify the performance of GRE methods; (2) human-annotated referential relations can be incomplete; (3) prompting LLMs with a fixed set of relations or entities can cause hallucinations. Next, we conducted a human evaluation of GRE methods that shows GENRES is consistent with human preferences for RE quality. Last, we made a comprehensive evaluation of fourteen leading LLMs using GENRES across document, bag, and sentence level RE datasets, respectively, to set the benchmark for future research in GRE. 

The advent of large language models (LLMs) has significantly advanced natural language processing tasks like text summarization. However, their large size and computational demands, coupled with privacy concerns in data transmission, limit their use in resourceconstrained and privacy-centric settings. To overcome this, we introduce TriSum, a framework for distilling LLMs’ text summarization abilities into a compact, local model. Initially, LLMs extract a set of aspect-triple rationales and summaries, which are refined using a dualscoring method for quality. Next, a smaller local model is trained with these tasks, employing a curriculum learning strategy that evolves from simple to complex tasks. Our method enhances local model performance on various benchmarks (CNN/DailyMail, XSum, and ClinicalTrial), outperforming baselines by 4.5%, 8.5%, and 7.4%, respectively. It also improves interpretability by providing insights into the summarization rationale. 

Roll-to-roll (R2R) printing techniques are promising for high-volume continuous production of substrate-based electronic products, as opposed to the sheet-to-sheet approach suited for low-volume work. However, one of the major challenges in R2R flexible electronics printing is achieving tight alignment tolerances, as specified by the device resolution (usually at micrometer level), for multi-layer printed electronics. The alignment of the printed patterns in different layers, known as registration, is critical to product quality. Registration errors are essentially accumulated positional or dimensional deviations caused by un-desired variations in web tensions and web speeds. Conventional registration control methods rely on model-based feedback controllers, such as PID control, to regulate the web tension and the web speed. However, those methods can not guarantee that the registration error always converges to zero due to lagging problems. In this paper, we propose a Spatial-Terminal Iterative Learning Control (STILC) method combined with PID control to enable the registration error to converge to zero iteratively, which achieves unprecedented control in the creation, integration and manipulation of multi-layer microstructures in R2R processes. We simulate the registration error generation and accumulation caused by axis mismatch between roller and motor that commonly exists in R2R systems. We show that the STILC-PID hybrid control method can eliminate the registration error completely after a reasonable number of iterations. We also compare the performances of STILC with a constant-value basis and a cosine-form basis. The results show that the control model with a cosine-form basis provides a faster convergence speed for R2R registration error elimination.

 

Roll-to-Roll (R2R) printing techniques are promising for high-volume continuous production of substrate-based products, as opposed to sheet-to-sheet (S2S) approach suited for low-volume work. However, meeting the tight alignment tolerance requirements of additive multi-layer printed electronics specified by device resolution that is usually at micrometer scale has become a major challenge in R2R flexible electronics printing, preventing the fabrication technology from being transferred from conventional S2S to high-speed R2R production. Print registration in a R2R process is to align successive print patterns on the flexible substrate and to ensure quality printed devices through effective control of various process variables. Conventional model-based control methods require an accurate web-handling dynamic model and real-time tension measurements to ensure control laws can be faithfully derived. For complex multistage R2R systems, physics-based state-space models are difficult to derive, and real-time tension measurements are not always acquirable. In this paper, we present a novel data-driven model predictive control (DD-MPC) method to minimize the multistage register errors effectively. We show that the DD-MPC can handle multi-input and multi-output systems and obtain the plant model from sensor data via an Eigensystem Realization Algorithm (ERA) and Observer Kalman filter identification (OKID) system identification method. In addition, the proposed control scheme works for systems with partially measurable system states.

 

Existing work in continual learning (CL) focuses on mitigating catastrophic forgetting, i.e., model performance deterioration on past tasks when learning a new task. However, the training efficiency of a CL system is under-investigated, which limits the real-world application of CL systems under resource-limited scenarios. In this work, we propose a novel framework called Sparse Continual Learning(SparCL), which is the first study that leverages sparsity to enable cost-effective continual learning on edge devices. SparCL achieves both training acceleration and accuracy preservation through the synergy of three aspects: weight sparsity, data efficiency, and gradient sparsity. Specifically, we propose task-aware dynamic masking (TDM) to learn a sparse network throughout the entire CL process, dynamic data removal (DDR) to remove less informative training data, and dynamic gradient masking (DGM) to sparsify the gradient updates. Each of them not only improves efficiency, but also further mitigates catastrophic forgetting. SparCL consistently improves the training efficiency of existing state-of-the-art (SOTA) CL methods by at most 23X less training FLOPs, and, surprisingly, further improves the SOTA accuracy by at most 1.7%. SparCL also outperforms competitive baselines obtained from adapting SOTA sparse training methods to the CL setting in both efficiency and accuracy. We also evaluate the effectiveness of SparCL on a real mobile phone, further indicating the practical potential of our method.