Search

The growth in prominence of large language models (LLMs) in everyday life can be largely attributed to their generative abilities, yet some of this is also owed to the risks and costs associated with their use. On one front is their tendency to \textit{hallucinate} false or misleading information, limiting their reliability. On another is the increasing focus on the computational limitations associated with traditional self-attention based LLMs, which has brought about new alternatives, in particular recurrent models, meant to overcome them. Yet it remains uncommon to consider these two concerns simultaneously. Do changes in architecture exacerbate/alleviate existing concerns about hallucinations? Do they affect how and where they occur? Through an extensive evaluation, we study how these architecture-based inductive biases affect the propensity to hallucinate. While hallucination remains a general phenomenon not limited to specific architectures, the situations in which they occur and the ease with which specific types of hallucinations can be induced can significantly differ based on the model architecture. These findings highlight the need for better understanding both these problems in conjunction with each other, as well as consider how to design more universal techniques for handling hallucinations.

We examine possible additive corrections to the Bekenstein-Hawking (BH) entropy of black holes due to very general classical and quantal modifications of general relativity. In general, black hole entropy is subject to the Generalized Second Law of Thermodynamics. For the case of binary black hole coalescence, the difference in corrections to the inspiral and remnant black hole entropies is shown, within this law, to be bounded by the difference in the corresponding BH entropies. This latter difference has been measured by several groups attempting to validate Hawking's Area Theorem on black hole horizons, by analyzing gravitational wave data from possible binary black hole mergers. The former difference - that between corrections to remnant and inspiral black hole entropies beyond the BH entropy, is thus constrained by a bound measured from observational data. We examine the implications of this constraint for general binary black hole coalesence. If calculated entropy corrections follow the essential pattern of BH entropies dictated by the Hawking Area Theorem, consistency with the observational bound is shown to be guaranteed. If they do not, these corrections are then nontrivially constrained by the observational bound., Comment: 5 pages Revtex 4-1, Abstract modified

Despite their widespread adoption, large language models (LLMs) remain prohibitive to use under resource constraints, with their ever growing sizes only increasing the barrier for use. One noted issue is the high latency associated with auto-regressive generation, rendering large LLMs use dependent on advanced computing infrastructure. Assisted decoding, where a smaller draft model guides a larger target model's generation, has helped alleviate this, but remains dependent on alignment between the two models. Thus if the draft model is insufficiently capable on some domain relative to the target model, performance can degrade. Alternatively, one can leverage multiple draft models to better cover the expertise of the target, but when multiple black-box draft models are available, selecting an assistant without details about its construction can be difficult. To better understand this decision making problem, we observe it as a contextual bandit, where a policy must choose a draft model based on a context. We show that even without prior knowledge of the draft models, creating an offline dataset from only outputs of independent draft/target models and training a policy over the alignment of these outputs can accelerate performance on multiple domains provided the candidates are effective. Further results show this to hold on various settings with multiple assisted decoding candidates, highlighting its flexibility and the advantageous role that such decision making can play., Comment: Published as a long paper at the 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP). Official version of paper within conference proceedings is available at http://aclanthology.org/2024.emnlp-main.332

Locations of DNA replication initiation in prokaryotes, called "origins of replication", are well-characterized. However, a mechanistic understanding of the sequence-dependence of the local unzipping of double-stranded DNA, the first step towards replication initiation, is lacking. Here, utilizing a Markov chain model that was created to address the directional nature of DNA unzipping and replication, we model the sequence dependence of local melting of double-stranded linear DNA segments. We show that generalized palindromic sequences with high nucleotide skews have a low kinetic barrier for local melting near melting temperatures. This allows for such sequences to function as replication origins. We support our claim with evidence for high-skew palindromic sequences within the replication origins of mitochondrial DNA, bacteria, archaea and plasmids., Comment: 20 pages, 9 figures

The emerging citation-based QA systems are gaining more attention especially in generative AI search applications. The importance of extracted knowledge provided to these systems is vital from both accuracy (completeness of information) and efficiency (extracting the information in a timely manner). In this regard, citation-based QA systems are suffering from two shortcomings. First, they usually rely only on web as a source of extracted knowledge and adding other external knowledge sources can hamper the efficiency of the system. Second, web-retrieved contents are usually obtained by some simple heuristics such as fixed length or breakpoints which might lead to splitting information into pieces. To mitigate these issues, we propose our enhanced web and efficient knowledge graph (KG) retrieval solution (EWEK-QA) to enrich the content of the extracted knowledge fed to the system. This has been done through designing an adaptive web retriever and incorporating KGs triples in an efficient manner. We demonstrate the effectiveness of EWEK-QA over the open-source state-of-the-art (SoTA) web-based and KG baseline models using a comprehensive set of quantitative and human evaluation experiments. Our model is able to: first, improve the web-retriever baseline in terms of extracting more relevant passages (>20\%), the coverage of answer span (>25\%) and self containment (>35\%); second, obtain and integrate KG triples into its pipeline very efficiently (by avoiding any LLM calls) to outperform the web-only and KG-only SoTA baselines significantly in 7 quantitative QA tasks and our human evaluation.

In this work, we dive deep into one of the popular knowledge-grounded dialogue benchmarks that focus on faithfulness, FaithDial. We show that a significant portion of the FaithDial data contains annotation artifacts, which may bias models towards completely ignoring the conversation history. We therefore introduce CHARP, a diagnostic test set, designed for an improved evaluation of hallucinations in conversational model. CHARP not only measures hallucination but also the compliance of the models to the conversation task. Our extensive analysis reveals that models primarily exhibit poor performance on CHARP due to their inability to effectively attend to and reason over the conversation history. Furthermore, the evaluation methods of FaithDial fail to capture these shortcomings, neglecting the conversational history. Our findings indicate that there is substantial room for contribution in both dataset creation and hallucination evaluation for knowledge-grounded dialogue, and that CHARP can serve as a tool for monitoring the progress in this particular research area. CHARP is publicly available at https://huggingface.co/datasets/huawei-noah/CHARP, Comment: To appear in Findings ACL 2024

Large language models (LLMs) show an innate skill for solving language based tasks. But insights have suggested an inability to adjust for information or task-solving skills becoming outdated, as their knowledge, stored directly within their parameters, remains static in time. Tool use helps by offloading work to systems that the LLM can access through an interface, but LLMs that use them still must adapt to nonstationary environments for prolonged use, as new tools can emerge and existing tools can change. Nevertheless, tools require less specialized knowledge, therefore we hypothesize they are better suited for continual learning (CL) as they rely less on parametric memory for solving tasks and instead focus on learning when to apply pre-defined tools. To verify this, we develop a synthetic benchmark and follow this by aggregating existing NLP tasks to form a more realistic testing scenario. While we demonstrate scaling model size is not a solution, regardless of tool usage, continual learning techniques can enable tool LLMs to both adapt faster while forgetting less, highlighting their potential as continual learners., Comment: 20 pages, 11 tables, 7 figures

This paper provides a few approaches to automating computer programming and project submission tasks, that we have been following for the last six years and have found to be successful. The approaches include using CodeRunner with Learning Management System (LMS) integration for programming practice and evaluation, and Git (GitHub) for project submissions and automatic code evaluation. In this paper, we describe the details of how we set up the tools and customized those for computer science courses. Based on our experiences, we also provide a few insights on using these tools for effective learning., Comment: 14 pages, 15 figures

Inspired by the pioneering 1968 work of L Parker, demonstrating matter quanta production in a dynamical spacetime background, we consider production of scalar quanta in a gravitational wave background. Choosing the spacetime to be a flat spacetime perturbed linearly by a linear gravitational wave, we show that scalar particles may indeed be produced in a perturbative manner. Our formulation is valid for any linear gravitational wave background profile, and is by no means restricted to monochromatic plane waves, in contrast to much of the earlier work on this topic. Thus, our work is directly applicable to gravitational wave signals from compact binary coalescence detected at LIGO, where they are of a pulsed character rather than monochromatic plane waves. We also briefly outline generalizing our approach for photon creation in a gravitational wave background. In this aspect, irrespective of the astrophysical nature of the binary merger sourcing the gravitational wave signal, one expects the dynamical nature of the spacetime to produce all species of light particles. Thus, any binary coalescence is in effect a source of multimessenger astrophysics., Comment: 6 pages revtex4-1, no figures, abstract changed, two sentences added. New paragraph added in Introduction to delineate the distinction of the paper from earlier work. Several new references added

Large Language Models (LLMs) have demonstrated superior performance in language understanding benchmarks. CALM, a popular approach, leverages linguistic priors of LLMs -- GPT-2 -- for action candidate recommendations to improve the performance in text games in Jericho without environment-provided actions. However, CALM adapts GPT-2 with annotated human gameplays and keeps the LLM fixed during the learning of the text based games. In this work, we explore and evaluate updating LLM used for candidate recommendation during the learning of the text based game as well to mitigate the reliance on the human annotated gameplays, which are costly to acquire. We observe that by updating the LLM during learning using carefully selected in-game transitions, we can reduce the dependency on using human annotated game plays for fine-tuning the LLMs. We conducted further analysis to study the transferability of the updated LLMs and observed that transferring in-game trained models to other games did not result in a consistent transfer.

In the age of artificial intelligence, the role of large language models (LLMs) is becoming increasingly central. Despite their growing prevalence, their capacity to consolidate knowledge from different training documents - a crucial ability in numerous applications - remains unexplored. This paper presents the first study examining the capability of LLMs to effectively combine such information within their parameter space. We introduce EpiK-Eval, a novel question-answering benchmark tailored to evaluate LLMs' proficiency in formulating a coherent and consistent knowledge representation from segmented narratives. Evaluations across various LLMs reveal significant weaknesses in this domain. We contend that these shortcomings stem from the intrinsic nature of prevailing training objectives. Consequently, we advocate for refining the approach towards knowledge consolidation, as it harbors the potential to dramatically improve their overall effectiveness and performance. The findings from this study offer insights for developing more robust and reliable LLMs. Our code and benchmark are available at https://github.com/chandar-lab/EpiK-Eval