Search

One of the prominent issues stifling the current generation of large language models is their limited context length. Recent proprietary models such as GPT-4 and Claude 2 have introduced longer context lengths, 8k/32k and 100k, respectively; however, despite the efforts in the community, most common models, such as LLama-2, have a context length of 4k or less. Unlimiformer (Bertsch et al., 2023) is a recently popular vector-retrieval augmentation method that offloads cross-attention computations to a kNN index. However, its main limitation is incompatibility with decoder-only transformers out of the box. In this work, we explore practical considerations of adapting Unlimiformer to decoder-only transformers and introduce a series of modifications to overcome this limitation. Moreover, we expand the original experimental setup on summarization to include a new task (i.e., free-form Q&A) and an instruction-tuned model (i.e., a custom 6.7B GPT model). Our results showcase the effectiveness of these modifications on summarization, performing on par with a model with 2x the context length. Moreover, we discuss limitations and future directions for free-form Q&A and instruction-tuned models., Comment: 8 pages, 6 figures

The COVID-19 pandemic profoundly impacted people globally, yet its effect on scientists and research institutions has yet to be fully examined. To address this knowledge gap, we use a newly available bibliographic dataset covering tens of millions of papers and authors to investigate changes in research activity and collaboration during this period. Employing statistical methods, we analyze the pandemic's disruptions on the participation, productivity, and collaborations of researchers at the top 1,000 institutions worldwide based on historical productivity, taking into account variables such as geography, researcher seniority and gender, and field of study. Our findings reveal an unexpected trend: research activity and output significantly increased in the early stages of the pandemic, indicating a surprising resilience in the scientific community. However, by the end of 2022, there was a notable reversion to historical trends in research participation and productivity. This reversion suggests that the initial spike in research activity was a short-lived disruption rather than a permanent shift. As such, monitoring scientific outputs in 2023 and beyond becomes crucial. There may be a delayed negative effect of the pandemic on research, given the long time horizon for many research fields and the temporary closure of wet labs. Further analysis is needed to fully comprehend the factors that underpin the resilience of scientific innovation in the face of global crises. Our study provides an initial comprehensive exploration up to the end of 2022, offering valuable insights into how the scientific community has adapted and responded over the course of the pandemic.

With the growing utilization of large language models (LLMs) across domains, alignment towards human preferences has become one of the most critical aspects of training models. At the forefront of state-of-the-art human alignment methods are preference optimization methods (*PO). However, prior research has often concentrated on identifying the best-performing method, typically involving a grid search over hyperparameters, which can be impractical for general practitioners. In this paper, we aim to identify the algorithm that, while being performant, is simultaneously more robust to varying hyperparameters, thereby increasing the likelihood of achieving better results. We focus on a realistic out-of-distribution (OOD) scenario that mirrors real-world applications of human alignment, offering practical insights into the strengths and weaknesses of these methods. Furthermore, to better understand the shortcomings of generations from the different methods, we analyze the model generations through the lens of KL divergence of the SFT model and the response length statistics. Our analysis reveals that the widely adopted DPO method consistently produces lengthy responses of inferior quality that are very close to the SFT responses. Motivated by these findings, we propose an embarrassingly simple extension to the DPO algorithm, LN-DPO, resulting in more concise responses without sacrificing quality compared to the policy obtained by vanilla DPO., Comment: 10 pages

While multi-modal large language models (MLLMs) have shown significant progress on many popular visual reasoning benchmarks, whether they possess abstract visual reasoning abilities remains an open question. Similar to the Sudoku puzzles, abstract visual reasoning (AVR) problems require finding high-level patterns (e.g., repetition constraints) that control the input shapes (e.g., digits) in a specific task configuration (e.g., matrix). However, existing AVR benchmarks only considered a limited set of patterns (addition, conjunction), input shapes (rectangle, square), and task configurations (3 by 3 matrices). To evaluate MLLMs' reasoning abilities comprehensively, we introduce MARVEL, a multidimensional AVR benchmark with 770 puzzles composed of six core knowledge patterns, geometric and abstract shapes, and five different task configurations. To inspect whether the model accuracy is grounded in perception and reasoning, MARVEL complements the general AVR question with perception questions in a hierarchical evaluation framework. We conduct comprehensive experiments on MARVEL with nine representative MLLMs in zero-shot and few-shot settings. Our experiments reveal that all models show near-random performance on the AVR question, with significant performance gaps (40%) compared to humans across all patterns and task configurations. Further analysis of perception questions reveals that MLLMs struggle to comprehend the visual features (near-random performance) and even count the panels in the puzzle ( <45%), hindering their ability for abstract reasoning. We release our entire code and dataset.

While large language models (LLMs) are still being adopted to new domains and utilized in novel applications, we are experiencing an influx of the new generation of foundation models, namely multi-modal large language models (MLLMs). These models integrate verbal and visual information, opening new possibilities to demonstrate more complex reasoning abilities at the intersection of the two modalities. However, despite the revolutionizing prospect of MLLMs, our understanding of their reasoning abilities is limited. In this study, we assess the nonverbal abstract reasoning abilities of open-source and closed-source MLLMs using variations of Raven's Progressive Matrices. Our experiments reveal the challenging nature of such problems for MLLMs while showcasing the immense gap between open-source and closed-source models. We also uncover critical shortcomings of visual and textual perceptions, subjecting the models to low-performance ceilings. Finally, to improve MLLMs' performance, we experiment with different methods, such as Chain-of-Thought prompting, leading to a significant (up to 100%) boost in performance. Our code and datasets are available at https://github.com/usc-isi-i2/isi-mmlm-rpm., Comment: 21 pages

Temporal knowledge graph (TKG) forecasting benchmarks challenge models to predict future facts using knowledge of past facts. In this paper, we apply large language models (LLMs) to these benchmarks using in-context learning (ICL). We investigate whether and to what extent LLMs can be used for TKG forecasting, especially without any fine-tuning or explicit modules for capturing structural and temporal information. For our experiments, we present a framework that converts relevant historical facts into prompts and generates ranked predictions using token probabilities. Surprisingly, we observe that LLMs, out-of-the-box, perform on par with state-of-the-art TKG models carefully designed and trained for TKG forecasting. Our extensive evaluation presents performances across several models and datasets with different characteristics, compares alternative heuristics for preparing contextual information, and contrasts to prominent TKG methods and simple frequency and recency baselines. We also discover that using numerical indices instead of entity/relation names, i.e., hiding semantic information, does not significantly affect the performance ($\pm$0.4\% Hit@1). This shows that prior semantic knowledge is unnecessary; instead, LLMs can leverage the existing patterns in the context to achieve such performance. Our analysis also reveals that ICL enables LLMs to learn irregular patterns from the historical context, going beyond simple predictions based on common or recent information., Comment: Accepted to EMNLP 2023 main conference. 14 pages, 4 figures, 10 tables

Research publications are the primary vehicle for sharing scientific progress in the form of new discoveries, methods, techniques, and insights. Unfortunately, the lack of a large-scale, comprehensive, and easy-to-use resource capturing the myriad relationships between publications, their authors, and venues presents a barrier to applications for gaining a deeper understanding of science. In this paper, we present PubGraph, a new resource for studying scientific progress that takes the form of a large-scale knowledge graph (KG) with more than 385M entities, 13B main edges, and 1.5B qualifier edges. PubGraph is comprehensive and unifies data from various sources, including Wikidata, OpenAlex, and Semantic Scholar, using the Wikidata ontology. Beyond the metadata available from these sources, PubGraph includes outputs from auxiliary community detection algorithms and large language models. To further support studies on reasoning over scientific networks, we create several large-scale benchmarks extracted from PubGraph for the core task of knowledge graph completion (KGC). These benchmarks present many challenges for knowledge graph embedding models, including an adversarial community-based KGC evaluation setting, zero-shot inductive learning, and large-scale learning. All of the aforementioned resources are accessible at https://pubgraph.isi.edu/ and released under the CC-BY-SA license. We plan to update PubGraph quarterly to accommodate the release of new publications., Comment: 17 Pages, 6 Figures, 9 Tables

Learning low-dimensional representations for entities and relations in knowledge graphs using contrastive estimation represents a scalable and effective method for inferring connectivity patterns. A crucial aspect of contrastive learning approaches is the choice of corruption distribution that generates hard negative samples, which force the embedding model to learn discriminative representations and find critical characteristics of observed data. While earlier methods either employ too simple corruption distributions, i.e. uniform, yielding easy uninformative negatives or sophisticated adversarial distributions with challenging optimization schemes, they do not explicitly incorporate known graph structure resulting in suboptimal negatives. In this paper, we propose Structure Aware Negative Sampling (SANS), an inexpensive negative sampling strategy that utilizes the rich graph structure by selecting negative samples from a node's k-hop neighborhood. Empirically, we demonstrate that SANS finds semantically meaningful negatives and is competitive with SOTA approaches while requires no additional parameters nor difficult adversarial optimization., Comment: Accepted to EMNLP 2020. Camera-ready submission

We present a multi-relational temporal Knowledge Graph based on the daily interactions between artifacts in GitHub, one of the largest social coding platforms. Such representation enables posing many user-activity and project management questions as link prediction and time queries over the knowledge graph. In particular, we introduce two new datasets for i) interpolated time-conditioned link prediction and ii) extrapolated time-conditioned link/time prediction queries, each with distinguished properties. Our experiments on these datasets highlight the potential of adapting knowledge graphs to answer broad software engineering questions. Meanwhile, it also reveals the unsatisfactory performance of existing temporal models on extrapolated queries and time prediction queries in general. To overcome these shortcomings, we introduce an extension to current temporal models using relative temporal information with regards to past events., Comment: 11 pages, 1 figure. 37th International Conference on Machine Learning (ICML 2020) - Workshop on Graph Representation Learning and Beyond

In this paper, we propose a novel writer-independent global feature extraction framework for the task of automatic signature verification which aims to make robust systems for automatically distinguishing negative and positive samples. Our method consists of an autoencoder for modeling the sample space into a fixed length latent space and a Siamese Network for classifying the fixed-length samples obtained from the autoencoder based on the reference samples of a subject as being "Genuine" or "Forged." During our experiments, usage of Attention Mechanism and applying Downsampling significantly improved the accuracy of the proposed framework. We evaluated our proposed framework using SigWiComp2013 Japanese and GPDSsyntheticOnLineOffLineSignature datasets. On the SigWiComp2013 Japanese dataset, we achieved 8.65% EER that means 1.2% relative improvement compared to the best-reported result. Furthermore, on the GPDSsyntheticOnLineOffLineSignature dataset, we achieved average EERs of 0.13%, 0.12%, 0.21% and 0.25% respectively for 150, 300, 1000 and 2000 test subjects which indicates improvement of relative EER on the best-reported result by 95.67%, 95.26%, 92.9% and 91.52% respectively. Apart from the accuracy gain, because of the nature of our proposed framework which is based on neural networks and consequently is as simple as some consecutive matrix multiplications, it has less computational cost than conventional methods such as DTW and could be used concurrently on devices such as GPU, TPU, etc., Comment: 13 pages, 10 figures, Submitted to Neural Computing and Applications journal