Search

In high-stakes domains like legal question-answering, the accuracy and trustworthiness of generative AI systems are of paramount importance. This work presents a comprehensive benchmark of various methods to assess the groundedness of AI-generated responses, aiming to significantly enhance their reliability. Our experiments include similarity-based metrics and natural language inference models to evaluate whether responses are well-founded in the given contexts. We also explore different prompting strategies for large language models to improve the detection of ungrounded responses. We validated the effectiveness of these methods using a newly created grounding classification corpus, designed specifically for legal queries and corresponding responses from retrieval-augmented prompting, focusing on their alignment with source material. Our results indicate potential in groundedness classification of generated responses, with the best method achieving a macro-F1 score of 0.8. Additionally, we evaluated the methods in terms of their latency to determine their suitability for real-world applications, as this step typically follows the generation process. This capability is essential for processes that may trigger additional manual verification or automated response regeneration. In summary, this study demonstrates the potential of various detection methods to improve the trustworthiness of generative AI in legal settings., Comment: to appear NLLP @ EMNLP 2024

Long-context large language models (LLMs) hold promise for tasks such as question-answering (QA) over long documents, but they tend to miss important information in the middle of context documents (arXiv:2307.03172v3). Here, we introduce $\textit{R&R}$ -- a combination of two novel prompt-based methods called $\textit{reprompting}$ and $\textit{in-context retrieval}$ (ICR) -- to alleviate this effect in document-based QA. In reprompting, we repeat the prompt instructions periodically throughout the context document to remind the LLM of its original task. In ICR, rather than instructing the LLM to answer the question directly, we instruct it to retrieve the top $k$ passage numbers most relevant to the given question, which are then used as an abbreviated context in a second QA prompt. We test R&R with GPT-4 Turbo and Claude-2.1 on documents up to 80k tokens in length and observe a 16-point boost in QA accuracy on average. Our further analysis suggests that R&R improves performance on long document-based QA because it reduces the distance between relevant context and the instructions. Finally, we show that compared to short-context chunkwise methods, R&R enables the use of larger chunks that cost fewer LLM calls and output tokens, while minimizing the drop in accuracy., Comment: 13 pages, 1 figure, 9 tables. For associated code repository see https://github.com/casetext/r-and-r

Most approaches for similar text retrieval and ranking with long natural language queries rely at some level on queries and responses having words in common with each other. Recent applications of transformer-based neural language models to text retrieval and ranking problems have been very promising, but still involve a two-step process in which result candidates are first obtained through bag-of-words-based approaches, and then reranked by a neural transformer. In this paper, we introduce novel approaches for effectively applying neural transformer models to similar text retrieval and ranking without an initial bag-of-words-based step. By eliminating the bag-of-words-based step, our approach is able to accurately retrieve and rank results even when they have no non-stopwords in common with the query. We accomplish this by using bidirectional encoder representations from transformers (BERT) to create vectorized representations of sentence-length texts, along with a vector nearest neighbor search index. We demonstrate both supervised and unsupervised means of using BERT to accomplish this task., Comment: 5 pages, 2 figures

Despite the prevalence of sputter deposition in the microelectronics industry, it has seen very limited applications for graphene electronics. In this letter, we report systematic investigation of the sputtering induced damages in graphene and identify the energetic sputtering gas neutrals as the primary cause of graphene disorder. We further demonstrate a grazing-incidence sputtering configuration that strongly suppresses fast neutral bombardment and retains graphene structure integrity, creating considerably lower damage than electron-beam evaporation. Such sputtering technique yields fully covered, smooth thin dielectric films, highlighting its potential for contact metals, gate oxides, and tunnel barriers fabrication in graphene device applications.

Multiferroics are singular materials that can display simultaneously electric and magnetic orders. Some of them can be ferroelectric and ferromagnetic and, for example, provide the unique opportunity of encoding information independently in electric polarization and magnetization to obtain four different logic states. However, schemes allowing a simple electrical readout of these different states have not been demonstrated so far. In this article, we show that this can be achieved if a multiferroic material is used as the tunnel barrier in a magnetic tunnel junction. We demonstrate that thin films of ferromagnetic-ferroelectric La0.1Bi0.9MnO3 (LBMO) retain both ferroic properties down to a thickness of only 2 nm. We have used such films as spin-filtering tunnel barriers the magnetization and electric polarization of which can be switched independently. In that case, the tunnel current across the structure is controlled by both the magnetic and ferroelectric configuration of the barrier, which gives rise to four distinct resistance states. This can be explained by the combination of spin filtering by the ferromagnetic LBMO barrier and the partial charge screening of electrical charges at the barrier/electrode interfaces due to ferroelectricity. We anticipate our results to be a starting point for more studies on the interplay between ferroelectricity and spin-dependent tunneling, and for the use of nanometric multiferroic elements in prototype devices. On a wider perspective, they may open the way towards novel reconfigurable logic spintronics architectures and to electrically controlled readout in quantum computing schemes using the spin-filter effect.