Selected Papers
Full list on Google Scholar.
Dr-CiK: A Testbed for Foresight-Driven Agents
Yihong Tang, Andrew Robert Williams, Arjun Ashok, Vincent Zhihao Zheng, Lijun Sun, Alexandre Drouin, Issam H. Laradji, Étienne Marcotte, Valentina Zantedeschi
Preprint
Dr-CiK benchmarks whether agents can find the right context to forecast the future: ground-truth evidence cuts error nearly 3x, yet today's best DR agents recover <5% of it and are fooled by distractors >80% of the time.
A benchmark for evaluating agents' ability to locate and utilize supporting context for time series forecasting. Rather than relying on pre-provided information, this testbed evaluates whether agents can independently discover relevant context from noisy sources. Findings reveal that most existing document retrieval agents recover only a small fraction of the ground-truth supporting evidence (<5%), and are frequently misled by distractors (>80% distractor citations), underscoring the need for more effective foresight-driven agents.
Overcoming the Modality Gap in Context-Aided Forecasting
Vincent Zhihao Zheng, Étienne Marcotte, Arjun Ashok, Andrew Robert Williams, Lijun Sun, Alexandre Drouin, Valentina Zantedeschi
Also at:
TSALM @ ICLR 2026 ·
Best Paper Award
A method to turn real-world time series datasets into multimodal context-aided forecasting datasets, which we use to generate a datasets containing 7 millions windows.
Context-aided forecasting (CAF) holds promise for integrating domain knowledge and forward-looking information, enabling AI systems to surpass traditional statistical methods. However, recent empirical studies reveal a puzzling gap: multimodal models often fail to outperform their unimodal counterparts. We hypothesize that this underperformance stems from poor context quality in existing datasets, as verification is challenging. To address these limitations, we introduce a semi-synthetic data augmentation method that generates contexts both descriptive of temporal dynamics and verifiably complementary to numerical histories. This approach enables massive-scale dataset creation, resulting in CAF-7M, a corpus of 7 million context-augmented time series windows, including a rigorously verified test set. We demonstrate that semi-synthetic pre-training transfers effectively to real-world evaluation, and show clear evidence of context utilization. Our results suggest that dataset quality, rather than architectural limitations, has been the primary bottleneck in context-aided forecasting.
Beyond Naïve Prompting: Strategies for Improved Context-aided Forecasting with LLMs
Arjun Ashok,
Andrew Robert Williams, Vincent Zhihao Zheng, Irina Rish, Nicolas Chapados, Étienne Marcotte, Valentina Zantedeschi, Alexandre Drouin
We introduce a unified framework for context-aided time series forecasting that bridges the LLM 'execution gap,' boosts prediction accuracy by up to 50%, and significantly reduces inference costs through adaptive model routing
Real-world forecasting requires models to integrate not only historical data but also relevant contextual information provided in textual form. While large language models (LLMs) show promise for context-aided forecasting, critical challenges remain: we lack diagnostic tools to understand failure modes, performance remains far below their potential, and high computational costs limit practical deployment. We introduce a unified framework of four strategies that address these limitations along three orthogonal dimensions: model diagnostics, accuracy, and efficiency. Through extensive evaluation across model families from small open-source models to frontier models including Gemini, GPT, and Claude, we uncover both fundamental insights and practical solutions. Our findings span three key dimensions: diagnostic strategies reveal the “Execution Gap” where models correctly explain how context affects forecasts but fail to apply this reasoning; accuracy-focused strategies achieve substantial performance improvements of 25-50%; and efficiency-oriented approaches show that adaptive routing between small and large models can approach large model accuracy on average while significantly reducing inference costs. These orthogonal strategies can be flexibly integrated based on deployment constraints, providing practitioners with a comprehensive toolkit for practical LLM-based context-aided forecasting.
Context is Key: A Benchmark for Forecasting with Essential Textual Information
Arjun Ashok*, Andrew Robert Williams*, Étienne Marcotte, Valentina Zantedeschi, Jithendaraa Subramanian, Roland Riachi, James Requeima, Alexandre Lacoste, Irina Rish, Nicolas Chapados, Alexandre Drouin (* co-first authorship)
A forecasting benchmark with problems that require the combined use of numerical historical data and textual context.
Forecasting is a critical task in decision making across various domains. While numerical data provides a foundation, it often lacks crucial context necessary for accurate predictions. Human forecasters frequently rely on additional information, such as background knowledge or constraints, which can be efficiently communicated through natural language. However, the ability of existing forecasting models to effectively integrate this textual information remains an open question. To address this, we introduce "Context is Key" (CiK), a time series forecasting benchmark that pairs numerical data with diverse types of carefully crafted textual context, requiring models to integrate both modalities. We evaluate a range of approaches, including statistical models, time series foundation models, and LLM-based forecasters, and propose a simple yet effective LLM prompting method that outperforms all other tested methods on our benchmark. Our experiments highlight the importance of incorporating contextual information, demonstrate surprising performance when using LLM-based forecasting models, and also reveal some of their critical shortcomings. By presenting this benchmark, we aim to advance multimodal forecasting, promoting models that are both accurate and accessible to decision-makers with varied technical expertise. The benchmark can be visualized at
https://servicenow.github.io/context-is-key-forecasting/.
TACTiS-2: Better, Faster, Simpler Attentional Copulas for Multivariate Time Series
Arjun Ashok, Étienne Marcotte, Valentina Zantedeschi, Nicolas Chapados, Alexandre Drouin
A flexible model for multivariate probabilistic time series prediction, simplifying the training of attentional copulas, with state-of-the-art accuracy on diverse forecasting tasks, while supporting interpolation and learning from irregular data.
We introduce a new model for multivariate probabilistic time series prediction, designed to flexibly address a range of tasks including forecasting, interpolation, and their combinations. Building on copula theory, we propose a simplified objective for the recently-introduced transformer-based attentional copulas (TACTiS), wherein the number of distributional parameters now scales linearly with the number of variables instead of factorially. The new objective requires the introduction of a training curriculum, which goes hand-in-hand with necessary changes to the original architecture. We show that the resulting model has significantly better training dynamics and achieves state-of-the-art performance across diverse real-world forecasting tasks, while maintaining the flexibility of prior work, such as seamless handling of unaligned and unevenly-sampled time series.
Lag-Llama: Towards Foundation Models for Time Series Forecasting
Arjun Ashok*, Kashif Rasul*, Andrew Robert Williams, Hena Ghonia, Rishika Bhagwatkar, Arian Khorasani, Mohammad Javad Darvishi Bayazi, George Adamopoulos, Roland Riachi, Nadhir Hassen, Marin Biloš, Sahil Garg, Anderson Schneider, Nicolas Chapados, Alexandre Drouin, Valentina Zantedeschi, Yuriy Nevmyvaka, Irina Rish (* co-first authorship)
Open-source model
A foundation model for probabilistic time series forecasting with strong zero-shot and few-shot capabilities
Over the past years, foundation models have caused a paradigm shift in machine learning due to their unprecedented capabilities for zero-shot and few-shot generalization. However, despite the success of foundation models in modalities such as natural language processing and computer vision, the development of foundation models for time series forecasting has lagged behind. We present Lag-Llama, a general-purpose foundation model for univariate probabilistic time series forecasting based on a decoder-only transformer architecture that uses lags as covariates. Lag-Llama is pretrained on a large corpus of diverse time series data from several domains, and demonstrates strong zero-shot generalization capabilities compared to a wide range of forecasting models on downstream datasets across domains. Moreover, when fine-tuned on relatively small fractions of such previously unseen datasets, Lag-Llama achieves state-of-the-art performance, outperforming prior deep learning approaches, emerging as the best general-purpose model on average. Lag-Llama serves as a strong contender to the current state-of-art in time series forecasting and paves the way for future advancements in foundation models tailored to time series data.
