Beyond Utility: Evaluating LLM as Recommender

With the rapid development of Large Language Models (LLMs), recent studies employed LLMs as recommenders to provide personalized information services for distinct users. Despite efforts to improve the accuracy of LLM-based recommendation models, relatively little attention is paid to beyond-utility dimensions. Moreover, there are unique evaluation aspects of LLM-based recommendation models, which have been largely ignored. To bridge this gap, we explore four new evaluation dimensions and propose a multidimensional evaluation framework. The new evaluation dimensions include: 1) history length sensitivity, 2) candidate position bias, 3) generation-involved performance, and 4) hallucinations. All four dimensions have the potential to impact performance, but are largely unnecessary for consideration in traditional systems. Using this multidimensional evaluation framework, along with traditional aspects, we evaluate the performance of seven LLM-based recommenders, with three prompting strategies, comparing them with six traditional models on both ranking and re-ranking tasks on four datasets. We find that LLMs excel at handling tasks with prior knowledge and shorter input histories in the ranking setting, and perform better in the re-ranking setting, beating traditional models across multiple dimensions. However, LLMs exhibit substantial candidate position bias issues, and some models hallucinate non-existent items much more often than others. We intend our evaluation framework and observations to benefit future research on the use of LLMs as recommenders. The code and data are available at https://github.com/JiangDeccc/EvaLLMasRecommender.

Report on the 1st Workshop on Large Language Model for Evaluation in Information Retrieval (LLM4Eval 2024) at SIGIR 2024

The first edition of the workshop on Large Language Model for Evaluation in Information Retrieval (LLM4Eval 2024) took place in July 2024, co-located with the ACM SIGIR Conference 2024 in the USA (SIGIR 2024). The aim was to bring information retrieval researchers together around the topic of LLMs for evaluation in information retrieval that gathered attention with the advancement of large language models and generative AI. Given the novelty of the topic, the workshop was focused around multi-sided discussions, namely panels and poster sessions of the accepted proceedings papers.

LLMJudge: LLMs for Relevance Judgments

The LLMJudge challenge is organized as part of the LLM4Eval workshop at SIGIR 2024. Test collections are essential for evaluating information retrieval (IR) systems. The evaluation and tuning of a search system is largely based on relevance labels, which indicate whether a document is useful for a specific search and user. However, collecting relevance judgments on a large scale is costly and resource-intensive. Consequently, typical experiments rely on third-party labelers who may not always produce accurate annotations. The LLMJudge challenge aims to explore an alternative approach by using LLMs to generate relevance judgments. Recent studies have shown that LLMs can generate reliable relevance judgments for search systems. However, it remains unclear which LLMs can match the accuracy of human labelers, which prompts are most effective, how fine-tuned open-source LLMs compare to closed-source LLMs like GPT-4, whether there are biases in synthetically generated data, and if data leakage affects the quality of generated labels. This challenge will investigate these questions, and the collected data will be released as a package to support automatic relevance judgment research in information retrieval and search.

Assessing and Verifying Task Utility in LLM-Powered Applications

The rapid development of Large Language Models (LLMs) has led to a surge in applications that facilitate collaboration among multiple agents, assisting humans in their daily tasks. However, a significant gap remains in assessing to what extent LLM-powered applications genuinely enhance user experience and task execution efficiency. This highlights the need to verify utility of LLM-powered applications, particularly by ensuring alignment between the application's functionality and end-user needs. We introduce AgentEval, a novel framework designed to simplify the utility verification process by automatically proposing a set of criteria tailored to the unique purpose of any given application. This allows for a comprehensive assessment, quantifying the utility of an application against the suggested criteria. We present a comprehensive analysis of the effectiveness and robustness of AgentEval for two open source datasets including Math Problem solving and ALFWorld House-hold related tasks. For reproducibility purposes, we make the data, code and all the logs publicly available at https://bit.ly/3w3yKcS .

Generative Information Retrieval Evaluation

In this chapter, we consider generative information retrieval evaluation from two distinct but interrelated perspectives. First, large language models (LLMs) themselves are rapidly becoming tools for evaluation, with current research indicating that LLMs may be superior to crowdsource workers and other paid assessors on basic relevance judgement tasks. We review past and ongoing related research, including speculation on the future of shared task initiatives, such as TREC, and a discussion on the continuing need for human assessments. Second, we consider the evaluation of emerging LLM-based generative information retrieval (GenIR) systems, including retrieval augmented generation (RAG) systems. We consider approaches that focus both on the end-to-end evaluation of GenIR systems and on the evaluation of a retrieval component as an element in a RAG system. Going forward, we expect the evaluation of GenIR systems to be at least partially based on LLM-based assessment, creating an apparent circularity, with a system seemingly evaluating its own output. We resolve this apparent circularity in two ways: 1) by viewing LLM-based assessment as a form of "slow search", where a slower IR system is used for evaluation and training of a faster production IR system; and 2) by recognizing a continuing need to ground evaluation in human assessment, even if the characteristics of that human assessment must change.

A Comparison of Methods for Evaluating Generative IR

Information retrieval systems increasingly incorporate generative components. For example, in a retrieval augmented generation (RAG) system, a retrieval component might provide a source of ground truth, while a generative component summarizes and augments its responses. In other systems, a large language model (LLM) might directly generate responses without consulting a retrieval component. While there are multiple definitions of generative information retrieval (Gen-IR) systems, in this paper we focus on those systems where the system's response is not drawn from a fixed collection of documents or passages. The response to a query may be entirely new text. Since traditional IR evaluation methods break down under this model, we explore various methods that extend traditional offline evaluation approaches to the Gen-IR context. Offline IR evaluation traditionally employs paid human assessors, but increasingly LLMs are replacing human assessment, demonstrating capabilities similar or superior to crowdsourced labels. Given that Gen-IR systems do not generate responses from a fixed set, we assume that methods for Gen-IR evaluation must largely depend on LLM-generated labels. Along with methods based on binary and graded relevance, we explore methods based on explicit subtopics, pairwise preferences, and embeddings. We first validate these methods against human assessments on several TREC Deep Learning Track tasks; we then apply these methods to evaluate the output of several purely generative systems. For each method we consider both its ability to act autonomously, without the need for human labels or other input, and its ability to support human auditing. To trust these methods, we must be assured that their results align with human assessments. In order to do so, evaluation criteria must be transparent, so that outcomes can be audited by human assessors.

Fréchet Distance for Offline Evaluation of Information Retrieval Systems with Sparse Labels

The rapid advancement of natural language processing, information retrieval (IR), computer vision, and other technologies has presented significant challenges in evaluating the performance of these systems. One of the main challenges is the scarcity of human-labeled data, which hinders the fair and accurate assessment of these systems. In this work, we specifically focus on evaluating IR systems with sparse labels, borrowing from recent research on evaluating computer vision tasks. taking inspiration from the success of using Fr\'echet Inception Distance (FID) in assessing text-to-image generation systems. We propose leveraging the Fr\'echet Distance to measure the distance between the distributions of relevant judged items and retrieved results. Our experimental results on MS MARCO V1 dataset and TREC Deep Learning Tracks query sets demonstrate the effectiveness of the Fr\'echet Distance as a metric for evaluating IR systems, particularly in settings where a few labels are available. This approach contributes to the advancement of evaluation methodologies in real-world scenarios such as the assessment of generative IR systems.

Adapting Standard Retrieval Benchmarks to Evaluate Generated Answers

Large language models can now directly generate answers to many factual questions without referencing external sources. Unfortunately, relatively little attention has been paid to methods for evaluating the quality and correctness of these answers, for comparing the performance of one model to another, or for comparing one prompt to another. In addition, the quality of generated answers are rarely directly compared to the quality of retrieved answers. As models evolve and prompts are modified, we have no systematic way to measure improvements without resorting to expensive human judgments. To address this problem we adapt standard retrieval benchmarks to evaluate answers generated by large language models. Inspired by the BERTScore metric for summarization, we explore two approaches. In the first, we base our evaluation on the benchmark relevance judgments. We empirically run experiments on how information retrieval relevance judgments can be utilized as an anchor to evaluating the generated answers. In the second, we compare generated answers to the top results retrieved by a diverse set of retrieval models, ranging from traditional approaches to advanced methods, allowing us to measure improvements without human judgments. In both cases, we measure the similarity between an embedded representation of the generated answer and an embedded representation of a known, or assumed, relevant passage from the retrieval benchmark.

Retrieving Supporting Evidence for Generative Question Answering

Current large language models (LLMs) can exhibit near-human levels of performance on many natural language-based tasks, including open-domain question answering. Unfortunately, at this time, they also convincingly hallucinate incorrect answers, so that responses to questions must be verified against external sources before they can be accepted at face value. In this paper, we report two simple experiments to automatically validate generated answers against a corpus. We base our experiments on questions and passages from the MS MARCO (V1) test collection, and a retrieval pipeline consisting of sparse retrieval, dense retrieval and neural rerankers. In the first experiment, we validate the generated answer in its entirety. After presenting a question to an LLM and receiving a generated answer, we query the corpus with the combination of the question + generated answer. We then present the LLM with the combination of the question + generated answer + retrieved answer, prompting it to indicate if the generated answer can be supported by the retrieved answer. In the second experiment, we consider the generated answer at a more granular level, prompting the LLM to extract a list of factual statements from the answer and verifying each statement separately. We query the corpus with each factual statement and then present the LLM with the statement and the corresponding retrieved evidence. The LLM is prompted to indicate if the statement can be supported and make necessary edits using the retrieved material. With an accuracy of over 80%, we find that an LLM is capable of verifying its generated answer when a corpus of supporting material is provided. However, manual assessment of a random sample of questions reveals that incorrect generated answers are missed by this verification process. While this verification process can reduce hallucinations, it can not entirely eliminate them.

Retrieving Supporting Evidence for LLMs Generated Answers

Current large language models (LLMs) can exhibit near-human levels of performance on many natural language tasks, including open-domain question answering. Unfortunately, they also convincingly hallucinate incorrect answers, so that responses to questions must be verified against external sources before they can be accepted at face value. In this paper, we report a simple experiment to automatically verify generated answers against a corpus. After presenting a question to an LLM and receiving a generated answer, we query the corpus with the combination of the question + generated answer. We then present the LLM with the combination of the question + generated answer + retrieved answer, prompting it to indicate if the generated answer can be supported by the retrieved answer. We base our experiment on questions and passages from the MS MARCO (V1) test collection, exploring three retrieval approaches ranging from standard BM25 to a full question answering stack, including a reader based on the LLM. For a large fraction of questions, we find that an LLM is capable of verifying its generated answer if appropriate supporting material is provided. However, with an accuracy of 70-80%, this approach cannot be fully relied upon to detect hallucinations.