Interleaving is an online evaluation approach for information retrieval systems that compares the effectiveness of ranking functions in interpreting the users' implicit feedback. Previous work such as Hofmann et al (2011) has evaluated the most promising interleaved methods at the time, on uniform distributions of queries. In the real world, ordinarily, there is an unbalanced distribution of repeated queries that follows a long-tailed users' search demand curve. The more a query is executed, by different users (or in different sessions), the higher the probability of collecting implicit feedback (interactions/clicks) on the related search results. This paper first aims to replicate the Team Draft Interleaving accuracy evaluation on uniform query distributions and then focuses on assessing how this method generalizes to long-tailed real-world scenarios. The reproducibility work raised interesting considerations on how the winning ranking function for each query should impact the overall winner for the entire evaluation. Based on what was observed, we propose that not all the queries should contribute to the final decision in equal proportion. As a result of these insights, we designed two variations of the $\Delta_{AB}$ score winner estimator that assign to each query a credit based on statistical hypothesis testing. To replicate, reproduce and extend the original work, we have developed from scratch a system that simulates a search engine and users' interactions from datasets from the industry. Our experiments confirm our intuition and show that our methods are promising in terms of accuracy, sensitivity, and robustness to noise.
相關內容
There is a rapidly growing number of large language models (LLMs) that users can query for a fee. We review the cost associated with querying popular LLM APIs, e.g. GPT-4, ChatGPT, J1-Jumbo, and find that these models have heterogeneous pricing structures, with fees that can differ by two orders of magnitude. In particular, using LLMs on large collections of queries and text can be expensive. Motivated by this, we outline and discuss three types of strategies that users can exploit to reduce the inference cost associated with using LLMs: 1) prompt adaptation, 2) LLM approximation, and 3) LLM cascade. As an example, we propose FrugalGPT, a simple yet flexible instantiation of LLM cascade which learns which combinations of LLMs to use for different queries in order to reduce cost and improve accuracy. Our experiments show that FrugalGPT can match the performance of the best individual LLM (e.g. GPT-4) with up to 98% cost reduction or improve the accuracy over GPT-4 by 4% with the same cost. The ideas and findings presented here lay a foundation for using LLMs sustainably and efficiently.
Pretrained Language Models (PLMs) have emerged as the state-of-the-art paradigm for code search tasks. The paradigm involves pretraining the model on search-irrelevant tasks such as masked language modeling, followed by the finetuning stage, which focuses on the search-relevant task. The typical finetuning method is to employ a dual-encoder architecture to encode semantic embeddings of query and code separately, and then calculate their similarity based on the embeddings. However, the typical dual-encoder architecture falls short in modeling token-level interactions between query and code, which limits the model's capabilities. In this paper, we propose a novel approach to address this limitation, introducing a cross-encoder architecture for code search that jointly encodes the semantic matching of query and code. We further introduce a Retriever-Ranker (RR) framework that cascades the dual-encoder and cross-encoder to promote the efficiency of evaluation and online serving. Moreover, we present a probabilistic hard negative sampling method to improve the cross-encoder's ability to distinguish hard negative codes, which further enhances the cascade RR framework. Experiments on four datasets using three code PLMs demonstrate the superiority of our proposed method.
Instruction tuning has been shown to be able to improve cross-task generalization of language models. However, it is still challenging for language models to complete the target tasks following the instructions, as the instructions are general and lack intermediate steps. To address this problem, we propose to incorporate the step-by-step instructions to help language models to decompose the tasks, which can provide the detailed and specific procedures for completing the target tasks. The step-by-step instructions are obtained automatically by prompting ChatGPT, which are further combined with the original instructions to tune language models. The extensive experiments on SUP-NATINST show that the high-quality step-by-step instructions can improve cross-task generalization across different model sizes. Moreover, the further analysis indicates the importance of the order of steps of the step-by-step instruction for the improvement. To facilitate future research, we release the step-by-step instructions and their human quality evaluation results.
Machine generated text is increasingly difficult to distinguish from human authored text. Powerful open-source models are freely available, and user-friendly tools that democratize access to generative models are proliferating. ChatGPT, which was released shortly after the first edition of this survey, epitomizes these trends. The great potential of state-of-the-art natural language generation (NLG) systems is tempered by the multitude of avenues for abuse. Detection of machine generated text is a key countermeasure for reducing abuse of NLG models, with significant technical challenges and numerous open problems. We provide a survey that includes both 1) an extensive analysis of threat models posed by contemporary NLG systems, and 2) the most complete review of machine generated text detection methods to date. This survey places machine generated text within its cybersecurity and social context, and provides strong guidance for future work addressing the most critical threat models, and ensuring detection systems themselves demonstrate trustworthiness through fairness, robustness, and accountability.
A circuit $\mathcal{C}$ samples a distribution $\mathbf{X}$ with an error $\epsilon$ if the statistical distance between the output of $\mathcal{C}$ on the uniform input and $\mathbf{X}$ is $\epsilon$. We study the hardness of sampling a uniform distribution over the set of $n$-bit strings of Hamming weight $k$ denoted by $\mathbf{U}^n_k$ for _decision forests_, i.e. every output bit is computed as a decision tree of the inputs. For every $k$ there is an $O(\log n)$-depth decision forest sampling $\mathbf{U}^n_k$ with an inverse-polynomial error [Viola 2012, Czumaj 2015]. We show that for every $\epsilon > 0$ there exists $\tau$ such that for decision depth $\tau \log (n/k) / \log \log (n/k)$, the error for sampling $\mathbf{U}_k^n$ is at least $1-\epsilon$. Our result is based on the recent robust sunflower lemma [Alweiss, Lovett, Wu, Zhang 2021, Rao 2019]. Our second result is about matching a set of $n$-bit strings with the image of a $d$-_local_ circuit, i.e. such that each output bit depends on at most $d$ input bits. We study the set of all $n$-bit strings whose Hamming weight is at least $n/2$. We improve the previously known locality lower bound from $\Omega(\log^* n)$ [Beyersdorff, Datta, Krebs, Mahajan, Scharfenberger-Fabian, Sreenivasaiah, Thomas and Vollmer, 2013] to $\Omega(\sqrt{\log n})$, leaving only a quartic gap from the best upper bound of $O(\log^2 n)$.
Profile likelihoods are rarely used in geostatistical models due to the computational burden imposed by repeated decompositions of large variance matrices. Accounting for uncertainty in covariance parameters can be highly consequential in geostatistical models as some covariance parameters are poorly identified, the problem is severe enough that the differentiability parameter of the Matern correlation function is typically treated as fixed. The problem is compounded with anisotropic spatial models as there are two additional parameters to consider. In this paper, we make the following contributions: 1, A methodology is created for profile likelihoods for Gaussian spatial models with Mat\'ern family of correlation functions, including anisotropic models. This methodology adopts a novel reparametrization for generation of representative points, and uses GPUs for parallel profile likelihoods computation in software implementation. 2, We show the profile likelihood of the Mat\'ern shape parameter is often quite flat but still identifiable, it can usually rule out very small values. 3, Simulation studies and applications on real data examples show that profile-based confidence intervals of covariance parameters and regression parameters have superior coverage to the traditional standard Wald type confidence intervals.
This research aims to explore various methods for assessing user feedback in mixed-initiative conversational search (CS) systems. While CS systems enjoy profuse advancements across multiple aspects, recent research fails to successfully incorporate feedback from the users. One of the main reasons for that is the lack of system-user conversational interaction data. To this end, we propose a user simulator-based framework for multi-turn interactions with a variety of mixed-initiative CS systems. Specifically, we develop a user simulator, dubbed ConvSim, that, once initialized with an information need description, is capable of providing feedback to a system's responses, as well as answering potential clarifying questions. Our experiments on a wide variety of state-of-the-art passage retrieval and neural re-ranking models show that effective utilization of user feedback can lead to 16% retrieval performance increase in terms of nDCG@3. Moreover, we observe consistent improvements as the number of feedback rounds increases (35% relative improvement in terms of nDCG@3 after three rounds). This points to a research gap in the development of specific feedback processing modules and opens a potential for significant advancements in CS. To support further research in the topic, we release over 30,000 transcripts of system-simulator interactions based on well-established CS datasets.
In a clinical trial with a survival outcome, an interim analysis is often performed to allow for early stopping for efficacy. If the interim analysis is early in the trial, one might conclude that a new treatment is more effective (compared to e.g.\ a placebo) and stop the trial, whereas the survival curves in the trial arms are not mature for the research question under investigation, for example because the curves are still close to 1 at that time. This means that the decision is based on a small percentage of the events in the long run only; possibly the events of the more frail patients in the trial who may not be representative for the whole group of patients. It may not be sensible to conclude effectiveness based on so little information. Criteria to determine the moment the interim analysis will be performed, should be chosen with care, and include the maturity of the data at the time of the interim analysis. Here, the expected survival rates at the interim analysis play a role. In this paper we will derive the asymptotic distribution of the Kaplan-Meier curves at the (random) moment the interim analysis will be performed for a one and two arm clinical trial. Based on this distribution, an interval in which the Kaplan Meier curves will fall into (with probability 95\%) is derived and could be used to plan the moment of the interim analysis in the design stage of the trial, so before the trial starts.
Results from Randomized Controlled Trials (RCTs) establish the comparative effectiveness of interventions, and are in turn critical inputs for evidence-based care. However, results from RCTs are presented in (often unstructured) natural language articles describing the design, execution, and outcomes of trials; clinicians must manually extract findings pertaining to interventions and outcomes of interest from such articles. This onerous manual process has motivated work on (semi-)automating extraction of structured evidence from trial reports. In this work we propose and evaluate a text-to-text model built on instruction-tuned Large Language Models (LLMs) to jointly extract Interventions, Outcomes, and Comparators (ICO elements) from clinical abstracts, and infer the associated results reported. Manual (expert) and automated evaluations indicate that framing evidence extraction as a conditional generation task and fine-tuning LLMs for this purpose realizes considerable ($\sim$20 point absolute F1 score) gains over the previous SOTA. We perform ablations and error analyses to assess aspects that contribute to model performance, and to highlight potential directions for further improvements. We apply our model to a collection of published RCTs through mid-2022, and release a searchable database of structured findings (anonymously for now): bit.ly/joint-relations-extraction-mlhc
Training a dialogue policy using deep reinforcement learning requires a lot of exploration of the environment. The amount of wasted invalid exploration makes their learning inefficient. In this paper, we find and define an important reason for the invalid exploration: dead-ends. When a conversation enters a dead-end state, regardless of the actions taken afterward, it will continue in a dead-end trajectory until the agent reaches a termination state or maximum turn. We propose a dead-end resurrection (DDR) algorithm that detects the initial dead-end state in a timely and efficient manner and provides a rescue action to guide and correct the exploration direction. To prevent dialogue policies from repeatedly making the same mistake, DDR also performs dialogue data augmentation by adding relevant experiences containing dead-end states. We first validate the dead-end detection reliability and then demonstrate the effectiveness and generality of the method by reporting experimental results on several dialogue datasets from different domains.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191