Search

Despite the striking advances in recent language generation performance, model-generated responses have suffered from the chronic problem of hallucinations that are either untrue or unfaithful to a given source. Especially in the task of knowledge grounded conversation, the models are required to generate informative responses, but hallucinated utterances lead to miscommunication. In particular, entity-level hallucination that causes critical misinformation and undesirable conversation is one of the major concerns. To address this issue, we propose a post-hoc refinement method called REM. It aims to enhance the quality and faithfulness of hallucinated utterances by refining them based on the source knowledge. If the generated utterance has a low source-faithfulness score with the given knowledge, REM mines the key entities in the knowledge and implicitly uses them for refining the utterances. We verify that our method reduces entity hallucination in the utterance. Also, we show the adaptability and efficacy of REM with extensive experiments and generative results. Our code is available at https://github.com/YOONNAJANG/REM., Comment: Accepted at EMNLP 2023

Minifilament (MF) eruption producing small jets and micro-flares is regarded as an important source for coronal heating and the solar wind transients through studies mostly based on coronal observations in the extreme ultraviolet (EUV) and X-ray wavelengths. In this study, we focus on the chromospheric plasma diagnostics of a tiny minifilament in quiet Sun located at [71'', 450''] on 2021--08--07 at 19:11 UT observed as part of the ninth encounter of the PSP campaign. Main data obtained are the high cadence, high resolution spectroscopy from the Fast Imaging Solar Spectrograph (FISS) and high-resolution magnetograms from the Near InfraRed Imaging Spectropolarimeter (NIRIS) on the 1.6~m Goode Solar Telescope (GST) at Big Bear Solar Observatory (BBSO). The mini-filament with size $\sim$1''$\times$5'' and a micro-flare are detected in both the H$\alpha$ line center and SDO/AIA 193, 304~\AA\ images. On the NIRIS magnetogram, we found that the cancellation of a magnetic bipole in the footpoints of the minifilament triggered its eruption in a sigmoidal shape. By inversion of the \ha\ and Ca {\sc ii} spectra under the embedded cloud model, we found a temperature increase of 3,800 K in the brightening region, associated with rising speed average of MF increased by 18~$km~s^{-1}$. This cool plasma is also found in the EUV images. We estimate the kinetic energy change of the rising filament as 1.5$\times$$10^{25}$~ergs, and thermal energy accumulation in the MF, 1.4$\times$$10^{25}$~ergs. From the photospheric magnetograms, we find the magnetic energy change is 1.6$\times$$10^{26}$~ergs across the PIL of converging opposite magnetic elements, which amounts to the energy release in the chromosphere in this smallest two-ribbon flare ever observed.

This work explores the zero-shot adaptation capability of semantic skills, semantically interpretable experts' behavior patterns, in cross-domain settings, where a user input in interleaved multi-modal snippets can prompt a new long-horizon task for different domains. In these cross-domain settings, we present a semantic skill translator framework SemTra which utilizes a set of multi-modal models to extract skills from the snippets, and leverages the reasoning capabilities of a pretrained language model to adapt these extracted skills to the target domain. The framework employs a two-level hierarchy for adaptation: task adaptation and skill adaptation. During task adaptation, seq-to-seq translation by the language model transforms the extracted skills into a semantic skill sequence, which is tailored to fit the cross-domain contexts. Skill adaptation focuses on optimizing each semantic skill for the target domain context, through parametric instantiations that are facilitated by language prompting and contrastive learning-based context inferences. This hierarchical adaptation empowers the framework to not only infer a complex task specification in one-shot from the interleaved multi-modal snippets, but also adapt it to new domains with zero-shot learning abilities. We evaluate our framework with Meta-World, Franka Kitchen, RLBench, and CARLA environments. The results clarify the framework's superiority in performing long-horizon tasks and adapting to different domains, showing its broad applicability in practical use cases, such as cognitive robots interpreting abstract instructions and autonomous vehicles operating under varied configurations., Comment: AAAI 2024 Camera-ready version