arXiv subjects

Daniel Campos

Publications and source records attributed to Daniel Campos.

At least 37 records · Page 2Linked to original sources

arXiv2024

A Large-Scale Study of Relevance Assessments with Large Language Models: An Initial Look

The application of large language models to provide relevance assessments presents exciting opportunities to advance information retrieval, natural language processing, and beyond, but to date many unknowns remain. This paper reports on the results of a large-scale evaluation (the TREC 2024 RAG Track) where four different relevance assessment approaches were deployed in situ: the "standard" fully manual process that NIST has implemented for decades and three different alternatives that take advantage of LLMs to different extents using the open-source UMBRELA tool. This setup allows us to correlate system rankings induced by the different approaches to characterize tradeoffs between cost and quality. We find that in terms of nDCG@20, nDCG@100, and Recall@100, system rankings induced by automatically generated relevance assessments from UMBRELA correlate highly with those induced by fully manual assessments across a diverse set of 77 runs from 19 teams. Our results suggest that automatically generated UMBRELA judgments can replace fully manual judgments to accurately capture run-level effectiveness. Surprisingly, we find that LLM assistance does not appear to increase correlation with fully manual assessments, suggesting that costs associated with human-in-the-loop processes do not bring obvious tangible benefits. Overall, human assessors appear to be stricter than UMBRELA in applying relevance criteria. Our work validates the use of LLMs in academic TREC-style evaluations and provides the foundation for future studies.

cs.IR↗

arXiv2024

Ragnarök: A Reusable RAG Framework and Baselines for TREC 2024 Retrieval-Augmented Generation Track

Did you try out the new Bing Search? Or maybe you fiddled around with Google AI~Overviews? These might sound familiar because the modern-day search stack has recently evolved to include retrieval-augmented generation (RAG) systems. They allow searching and incorporating real-time data into large language models (LLMs) to provide a well-informed, attributed, concise summary in contrast to the traditional search paradigm that relies on displaying a ranked list of documents. Therefore, given these recent advancements, it is crucial to have an arena to build, test, visualize, and systematically evaluate RAG-based search systems. With this in mind, we propose the TREC 2024 RAG Track to foster innovation in evaluating RAG systems. In our work, we lay out the steps we've made towards making this track a reality -- we describe the details of our reusable framework, Ragnarök, explain the curation of the new MS MARCO V2.1 collection choice, release the development topics for the track, and standardize the I/O definitions which assist the end user. Next, using Ragnarök, we identify and provide key industrial baselines such as OpenAI's GPT-4o or Cohere's Command R+. Further, we introduce a web-based user interface for an interactive arena allowing benchmarking pairwise RAG systems by crowdsourcing. We open-source our Ragnarök framework and baselines to achieve a unified standard for future RAG systems.

cs.IR↗

arXiv2024

Synthetic Test Collections for Retrieval Evaluation

Test collections play a vital role in evaluation of information retrieval (IR) systems. Obtaining a diverse set of user queries for test collection construction can be challenging, and acquiring relevance judgments, which indicate the appropriateness of retrieved documents to a query, is often costly and resource-intensive. Generating synthetic datasets using Large Language Models (LLMs) has recently gained significant attention in various applications. In IR, while previous work exploited the capabilities of LLMs to generate synthetic queries or documents to augment training data and improve the performance of ranking models, using LLMs for constructing synthetic test collections is relatively unexplored. Previous studies demonstrate that LLMs have the potential to generate synthetic relevance judgments for use in the evaluation of IR systems. In this paper, we comprehensively investigate whether it is possible to use LLMs to construct fully synthetic test collections by generating not only synthetic judgments but also synthetic queries. In particular, we analyse whether it is possible to construct reliable synthetic test collections and the potential risks of bias such test collections may exhibit towards LLM-based models. Our experiments indicate that using LLMs it is possible to construct synthetic test collections that can reliably be used for retrieval evaluation.

cs.IR↗

arXiv2024

Arctic-Embed: Scalable, Efficient, and Accurate Text Embedding Models

This report describes the training dataset creation and recipe behind the family of \texttt{arctic-embed} text embedding models (a set of five models ranging from 22 to 334 million parameters with weights open-sourced under an Apache-2 license). At the time of their release, each model achieved state-of-the-art retrieval accuracy for models of their size on the MTEB Retrieval leaderboard, with the largest model, arctic-embed-l outperforming closed source embedding models such as Cohere's embed-v3 and Open AI's text-embed-3-large. In addition to the details of our training recipe, we have provided several informative ablation studies, which we believe are the cause of our model performance.

cs.CL↗

arXiv2024

Intermittent random walks under stochastic resetting

We analyze a one-dimensional intermittent random walk on an unbounded domain in the presence of stochastic resetting. In this process, the walker alternates between local intensive search, diffusion, and rapid ballistic relocations in which it does not react to the target. We demonstrate that Poissonian resetting leads to the existence of a non-equilibrium steady state. We calculate the distribution of the first arrival time to a target along with its mean and show the existence of an optimal reset rate. In particular, we prove that the initial condition of the walker, i.e., either starting diffusely or relocating, can significantly affect the long-time properties of the search process. Moreover, we demonstrate the presence of distinct parameter regimes for the global optimization of the mean first arrival time when ballistic and diffusive movements are in direct competition.

cond-mat.stat-mech↗

arXiv2024

First-passage time of a Brownian searcher with stochastic resetting to random positions

We study the effect of a resetting point randomly distributed around the origin on the mean first passage time of a Brownian searcher moving in one dimension. We compare the search efficiency with that corresponding to reset to the origin and find that the mean first passage time of the latter can be larger or smaller than the distributed case, depending on whether the resetting points are symmetrically or asymmetrically distributed. In particular, we prove the existence of an optimal reset rate that minimizes the mean first-passage time for distributed resetting to a finite interval if the target is located outside this interval. When the target position belongs to the resetting interval or it is infinite then no optimal reset rate exists, but there is an optimal resetting interval width or resetting characteristic scale which minimizes the mean first-passage time. We also show that the first-passage density averaged over the resetting points depends on its first moment only. As a consequence, there is an equivalent point such that the first-passage problem with resetting to that point is statistically equivalent to the case of distributed resetting. We end our study by analyzing the fluctuations of the first-passage times for these cases. All our analytical results are verified through numerical simulations.

cond-mat.stat-mech↗

arXiv2023

Dynamic redundancy as a mechanism to optimize collective random searches

We explore the case of a group of random walkers looking for a target randomly located in space, such that the number of walkers is not constant but new ones can join the search, or those that are active can abandon it, with constant rates $r_b$ and $r_d$, respectively. Exact analytical solutions are provided both for the fastest-first-passage time and for the collective search time required to reach the target, in the seminal case of Brownian walkers with $r_d=0$. We prove that even for such a simplified situation there exists an optimal rate $r_b$ at which walkers should join the search to minimize the collective effort required to reach the target. We discuss how these results open a new line to understand the optimal regulation of cooperative random searches, e.g. for the case of biological foraging in social species.

cond-mat.stat-mech↗

arXiv2023

Overview of the TREC 2023 Product Product Search Track

This is the first year of the TREC Product search track. The focus this year was the creation of a reusable collection and evaluation of the impact of the use of metadata and multi-modal data on retrieval accuracy. This year we leverage the new product search corpus, which includes contextual metadata. Our analysis shows that in the product search domain, traditional retrieval systems are highly effective and commonly outperform general-purpose pretrained embedding models. Our analysis also evaluates the impact of using simplified and metadata-enhanced collections, finding no clear trend in the impact of the expanded collection. We also see some surprising outcomes; despite their widespread adoption and competitive performance on other tasks, we find single-stage dense retrieval runs can commonly be noncompetitive or generate low-quality results both in the zero-shot and fine-tuned domain.

cs.IR↗

arXiv2023

To Asymmetry and Beyond: Structured Pruning of Sequence to Sequence Models for Improved Inference Efficiency

Sequence-to-sequence language models can be used to produce abstractive summaries which are coherent, relevant, and concise. Still, model sizes can make deployment in latency-sensitive or web-scale implementations difficult. This paper studies the relationship between model size, structured pruning, inference efficiency, and summarization accuracy on widely used summarization datasets. We show that model accuracy is tied to the encoder size while inference efficiency is connected to the decoder. Using asymmetric pruning can lead to nearly 3x improvement in inference latency with ~1 point loss in Rouge-2. Moreover, we find both the average degradation and the role of asymmetry to be consistent across model sizes and variations in datasets.

cs.CL↗

arXiv2023

oBERTa: Improving Sparse Transfer Learning via improved initialization, distillation, and pruning regimes

In this paper, we introduce the range of oBERTa language models, an easy-to-use set of language models which allows Natural Language Processing (NLP) practitioners to obtain between 3.8 and 24.3 times faster models without expertise in model compression. Specifically, oBERTa extends existing work on pruning, knowledge distillation, and quantization and leverages frozen embeddings improves distillation and model initialization to deliver higher accuracy on a broad range of transfer tasks. In generating oBERTa, we explore how the highly optimized RoBERTa differs from the BERT for pruning during pre-training and finetuning. We find it less amenable to compression during fine-tuning. We explore the use of oBERTa on seven representative NLP tasks and find that the improved compression techniques allow a pruned oBERTa model to match the performance of BERTbase and exceed the performance of Prune OFA Large on the SQUAD V1.1 Question Answering dataset, despite being 8x and 2x, respectively faster in inference. We release our code, training regimes, and associated model for broad usage to encourage usage and experimentation

cs.CL↗

arXiv2023

Quick Dense Retrievers Consume KALE: Post Training Kullback Leibler Alignment of Embeddings for Asymmetrical dual encoders

In this paper, we consider the problem of improving the inference latency of language model-based dense retrieval systems by introducing structural compression and model size asymmetry between the context and query encoders. First, we investigate the impact of pre and post-training compression on the MSMARCO, Natural Questions, TriviaQA, SQUAD, and SCIFACT, finding that asymmetry in the dual encoders in dense retrieval can lead to improved inference efficiency. Knowing this, we introduce Kullback Leibler Alignment of Embeddings (KALE), an efficient and accurate method for increasing the inference efficiency of dense retrieval methods by pruning and aligning the query encoder after training. Specifically, KALE extends traditional Knowledge Distillation after bi-encoder training, allowing for effective query encoder compression without full retraining or index generation. Using KALE and asymmetric training, we can generate models which exceed the performance of DistilBERT despite having 3x faster inference.

cs.CL↗

arXiv2023

Noise-Robust Dense Retrieval via Contrastive Alignment Post Training

The success of contextual word representations and advances in neural information retrieval have made dense vector-based retrieval a standard approach for passage and document ranking. While effective and efficient, dual-encoders are brittle to variations in query distributions and noisy queries. Data augmentation can make models more robust but introduces overhead to training set generation and requires retraining and index regeneration. We present Contrastive Alignment POst Training (CAPOT), a highly efficient finetuning method that improves model robustness without requiring index regeneration, the training set optimization, or alteration. CAPOT enables robust retrieval by freezing the document encoder while the query encoder learns to align noisy queries with their unaltered root. We evaluate CAPOT noisy variants of MSMARCO, Natural Questions, and Trivia QA passage retrieval, finding CAPOT has a similar impact as data augmentation with none of its overhead.

cs.IR↗

arXiv2023

Sparse*BERT: Sparse Models Generalize To New tasks and Domains

Large Language Models have become the core architecture upon which most modern natural language processing (NLP) systems build. These models can consistently deliver impressive accuracy and robustness across tasks and domains, but their high computational overhead can make inference difficult and expensive. To make using these models less costly, recent work has explored leveraging structured and unstructured pruning, quantization, and distillation to improve inference speed and decrease size. This paper studies how models pruned using Gradual Unstructured Magnitude Pruning can transfer between domains and tasks. Our experimentation shows that models that are pruned during pretraining using general domain masked language models can transfer to novel domains and tasks without extensive hyperparameter exploration or specialized approaches. We demonstrate that our general sparse model Sparse*BERT can become SparseBioBERT simply by pretraining the compressed architecture on unstructured biomedical text. Moreover, we show that SparseBioBERT can match the quality of BioBERT with only 10\% of the parameters.

cs.CL↗

arXiv2023

Dense Sparse Retrieval: Using Sparse Language Models for Inference Efficient Dense Retrieval

Vector-based retrieval systems have become a common staple for academic and industrial search applications because they provide a simple and scalable way of extending the search to leverage contextual representations for documents and queries. As these vector-based systems rely on contextual language models, their usage commonly requires GPUs, which can be expensive and difficult to manage. Given recent advances in introducing sparsity into language models for improved inference efficiency, in this paper, we study how sparse language models can be used for dense retrieval to improve inference efficiency. Using the popular retrieval library Tevatron and the MSMARCO, NQ, and TriviaQA datasets, we find that sparse language models can be used as direct replacements with little to no drop in accuracy and up to 4.3x improved inference speeds

cs.IR↗

arXiv2022

Compressing Cross-Lingual Multi-Task Models at Qualtrics

Experience management is an emerging business area where organizations focus on understanding the feedback of customers and employees in order to improve their end-to-end experiences. This results in a unique set of machine learning problems to help understand how people feel, discover issues they care about, and find which actions need to be taken on data that are different in content and distribution from traditional NLP domains. In this paper, we present a case study of building text analysis applications that perform multiple classification tasks efficiently in 12 languages in the nascent business area of experience management. In order to scale up modern ML methods on experience data, we leverage cross lingual and multi-task modeling techniques to consolidate our models into a single deployment to avoid overhead. We also make use of model compression and model distillation to reduce overall inference latency and hardware cost to the level acceptable for business needs while maintaining model prediction quality. Our findings show that multi-task modeling improves task performance for a subset of experience management tasks in both XLM-R and mBert architectures. Among the compressed architectures we explored, we found that MiniLM achieved the best compression/performance tradeoff. Our case study demonstrates a speedup of up to 15.61x with 2.60% average task degradation (or 3.29x speedup with 1.71% degradation) and estimated savings of 44% over using the original full-size model. These results demonstrate a successful scaling up of text classification for the challenging new area of ML for experience management.

cs.CL↗

arXiv2022

The Optimal BERT Surgeon: Scalable and Accurate Second-Order Pruning for Large Language Models

Transformer-based language models have become a key building block for natural language processing. While these models are extremely accurate, they can be too large and computationally intensive to run on standard deployments. A variety of compression methods, including distillation, quantization, structured and unstructured pruning are known to decrease model size and increase inference speed, with low accuracy loss. In this context, this paper's contributions are two-fold. We perform an in-depth study of the accuracy-compression trade-off for unstructured weight pruning of BERT models. We introduce Optimal BERT Surgeon (oBERT), an efficient and accurate weight pruning method based on approximate second-order information, which we show to yield state-of-the-art results in both stages of language tasks: pre-training and fine-tuning. Specifically, oBERT extends existing work on unstructured second-order pruning by allowing for pruning blocks of weights, and by being applicable at the BERT scale. Second, we investigate the impact of this pruning method when compounding compression approaches to obtain highly compressed but accurate models for deployment on edge devices. These models significantly push boundaries of the current state-of-the-art sparse BERT models with respect to all metrics: model size, inference speed and task accuracy. For example, relative to the dense BERT-base, we obtain 10x model size compression (in MB) with < 1% accuracy drop, 10x CPU-inference speedup with < 2% accuracy drop, and 29x CPU-inference speedup with < 7.5% accuracy drop. Our code, fully integrated with Transformers and SparseML, is available at https://github.com/neuralmagic/sparseml/tree/main/research/optimal_BERT_surgeon_oBERT.

cs.CL↗

arXiv2022

Conditioned backward and forward times of diffusion with stochastic resetting: a renewal theory approach

Stochastic resetting can be naturally understood as a renewal process governing the evolution of an underlying stochastic process. In this work, we formally derive well-known results of diffusion with resets from a renewal theory perspective. Parallel to the concepts from renewal theory, we introduce the conditioned backward and forward times for stochastic processes with resetting to be the times since the last and until the next reset, given that the current state of the system is known. We focus on studying diffusion under Markovian and non-Markovian resetting. For these cases, we find the conditioned backward and forward time PDFs, comparing them with numerical simulations of the process. In particular, we find that for power-law reset time PDFs with asymptotic form $φ(t)\sim t^{-1-α}$, significant changes in the properties of the conditioned backward and forward times happen at half-integer values of $α$. This is due to the composition between the long-time scaling of diffusion $P(x,t)\sim 1/\sqrt{t}$ and the reset time PDF.

cond-mat.stat-mech↗

arXiv2022

Non-standard diffusion under Markovian resetting in bounded domains

We consider a walker moving in a one-dimensional interval with absorbing boundaries under the effect of Markovian resettings to the initial position. The walker's motion follows a random walk characterized by a general waiting time distribution between consecutive short jumps. We investigate the existence of an optimal reset rate, which minimizes the mean exit passage time, in terms of the statistical properties of the waiting time probability. Generalizing previous results restricted to Markovian random walks, we here find that, depending on the value of the relative standard deviation of the waiting time probability, resetting can be either (i) never beneficial, (ii) beneficial depending on the distance of the reset to the boundary, or (iii) always beneficial.

cond-mat.stat-mech↗