TFRBench: A Reasoning Benchmark for Evaluating Forecasting Systems

Md Atik Ahamed, Mihir Parmar, Palash Goyal, Yiwen Song, Long T. Le, Qiang Cheng, Chun-Liang Li, Hamid Palangi, Jinsung Yoon, Tomas Pfister
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:1169-1223, 2026.

Abstract

We introduce TFRBench, the first benchmark designed to evaluate the reasoning capabilities of forecasting systems. Traditionally, time-series forecasting has been evaluated solely on numerical accuracy, treating foundation models as "black boxes." Unlike existing benchmarks, TFRBench provides a protocol for evaluating the reasoning generated by forecasting systems–specifically their analysis of cross-channel dependencies, trends, and external events. To enable this, we propose a systematic multi-agent framework that utilizes an iterative verification loop to synthesize numerically grounded reasoning traces. Spanning ten datasets across five domains, our evaluation confirms that this reasoning is causally effective; useful for evaluation; and prompting LLMs with our generated traces significantly improves forecasting accuracy compared to direct numerical prediction (e.g., avg. $\sim40.2$% $\rightarrow$ $\sim56.6$%), validating the quality of our reasoning. Conversely, benchmarking experiments reveal that off-the-shelf LLMs consistently struggle with both reasoning (lower LLM-as-a-Judge scores) and numerical forecasting, frequently failing to capture domain-specific dynamics. TFRBench thus establishes a new standard for interpretable, reasoning-based evaluation in time-series forecasting. Our benchmark is available at: https://tfrbench.github.io

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-ahamed26a, title = {{TFRB}ench: A Reasoning Benchmark for Evaluating Forecasting Systems}, author = {Ahamed, Md Atik and Parmar, Mihir and Goyal, Palash and Song, Yiwen and Le, Long T. and Cheng, Qiang and Li, Chun-Liang and Palangi, Hamid and Yoon, Jinsung and Pfister, Tomas}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {1169--1223}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/ahamed26a/ahamed26a.pdf}, url = {https://proceedings.mlr.press/v306/ahamed26a.html}, abstract = {We introduce TFRBench, the first benchmark designed to evaluate the reasoning capabilities of forecasting systems. Traditionally, time-series forecasting has been evaluated solely on numerical accuracy, treating foundation models as "black boxes." Unlike existing benchmarks, TFRBench provides a protocol for evaluating the reasoning generated by forecasting systems–specifically their analysis of cross-channel dependencies, trends, and external events. To enable this, we propose a systematic multi-agent framework that utilizes an iterative verification loop to synthesize numerically grounded reasoning traces. Spanning ten datasets across five domains, our evaluation confirms that this reasoning is causally effective; useful for evaluation; and prompting LLMs with our generated traces significantly improves forecasting accuracy compared to direct numerical prediction (e.g., avg. $\sim40.2$% $\rightarrow$ $\sim56.6$%), validating the quality of our reasoning. Conversely, benchmarking experiments reveal that off-the-shelf LLMs consistently struggle with both reasoning (lower LLM-as-a-Judge scores) and numerical forecasting, frequently failing to capture domain-specific dynamics. TFRBench thus establishes a new standard for interpretable, reasoning-based evaluation in time-series forecasting. Our benchmark is available at: https://tfrbench.github.io} }
Endnote
%0 Conference Paper %T TFRBench: A Reasoning Benchmark for Evaluating Forecasting Systems %A Md Atik Ahamed %A Mihir Parmar %A Palash Goyal %A Yiwen Song %A Long T. Le %A Qiang Cheng %A Chun-Liang Li %A Hamid Palangi %A Jinsung Yoon %A Tomas Pfister %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-ahamed26a %I PMLR %P 1169--1223 %U https://proceedings.mlr.press/v306/ahamed26a.html %V 306 %X We introduce TFRBench, the first benchmark designed to evaluate the reasoning capabilities of forecasting systems. Traditionally, time-series forecasting has been evaluated solely on numerical accuracy, treating foundation models as "black boxes." Unlike existing benchmarks, TFRBench provides a protocol for evaluating the reasoning generated by forecasting systems–specifically their analysis of cross-channel dependencies, trends, and external events. To enable this, we propose a systematic multi-agent framework that utilizes an iterative verification loop to synthesize numerically grounded reasoning traces. Spanning ten datasets across five domains, our evaluation confirms that this reasoning is causally effective; useful for evaluation; and prompting LLMs with our generated traces significantly improves forecasting accuracy compared to direct numerical prediction (e.g., avg. $\sim40.2$% $\rightarrow$ $\sim56.6$%), validating the quality of our reasoning. Conversely, benchmarking experiments reveal that off-the-shelf LLMs consistently struggle with both reasoning (lower LLM-as-a-Judge scores) and numerical forecasting, frequently failing to capture domain-specific dynamics. TFRBench thus establishes a new standard for interpretable, reasoning-based evaluation in time-series forecasting. Our benchmark is available at: https://tfrbench.github.io
APA
Ahamed, M.A., Parmar, M., Goyal, P., Song, Y., Le, L.T., Cheng, Q., Li, C., Palangi, H., Yoon, J. & Pfister, T.. (2026). TFRBench: A Reasoning Benchmark for Evaluating Forecasting Systems. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:1169-1223 Available from https://proceedings.mlr.press/v306/ahamed26a.html.

Related Material