Search for: All records

Toxic content detection is crucial for online services to remove inappropriate content that violates community standards. To automate the detection process, prior works have proposed varieties of machine learning (ML) approaches to train Language Models (LMs) for toxic content detection. However, both their accuracy and transferability across datasets are limited. Recently, Large Language Models (LLMs) have shown promise in toxic content detection due to their superior zero-shot and few-shot in-context learning ability as well as broad transferability on ML tasks.However, efficiently designing prompts for LLMs remains challenging. Moreover, the high run-time cost of LLMs may hinder their deployments in production. To address these challenges, in this work, we propose BD-LLM, a novel and efficient approach to bootstrapping and distilling LLMs for toxic content detection. Specifically, we design a novel prompting method named Decision-Tree-of-Thought (DToT) to bootstrap LLMs' detection performance and extract high-quality rationales. DToT can automatically select more fine-grained context to re-prompt LLMs when their responses lack confidence. Additionally, we use the rationales extracted via DToT to fine-tune student LMs. Our experimental results on various datasets demonstrate that DToT can improve the accuracy of LLMs by up to 4.6%. Furthermore, student LMs fine-tuned with rationales extracted via DToT outperform baselines on all datasets with up to 16.9% accuracy improvement, while being more than 60x smaller than conventional LLMs. Finally, we observe that student LMs fine-tuned with rationales exhibit better cross-dataset transferability. 

Abstract Permafrost dynamics can drastically affect vegetation and soil carbon dynamics in northern high latitudes. Vegetation has significant influences on the energy balance of soil surface by impacting the short-wave radiation, long-wave radiation and surface sensible heat flux, affecting soil thermal dynamics, in turn, inducing vegetation shift, affecting carbon cycling. During winter, snow can also significantly impact soil temperature due to its insulative effect. However, these processes have not been fully modeled to date. To quantify the interactions between vegetation, snow, and soil thermal dynamics and their impacts on carbon dynamics over the circumpolar region (45–90° N), we revise a sophisticated ecosystem model to improve simulations of soil temperature profile and their influences on vegetation, ecosystem carbon pools and fluxes. We find that, with warmer soil temperature in winter and cooler soil temperature in summer simulated with the revised model considering vegetation shift and snow effects, the region will release 1.54 Pg C/year to the atmosphere for present-day and 66.77–87.95 Pg C in 2022–2100. The canopy effects due to vegetation shift, however, will get more carbon sequestered into the ecosystem at 1.00 Pg C/year for present day and 36.09–44.32 Pg C/year in 2022–2100. This study highlights the importance to consider the interactions between snow, vegetation shift and soil thermal dynamics in simulating carbon dynamics in the region.