From 05704d0eb2fa59fe727652465b07db40bcb06c38 Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Wed, 8 Apr 2026 22:06:05 -0500 Subject: Initial release: GAP framework - Full pipeline: variant generation, multi-judge verification, evaluation - Loaders for OpenAI / Anthropic / Google / xAI / OpenRouter / vLLM - Framework-level mechanism analyses: paired structural overlap, repairability rescue, self-correction probe, cross-model agreement, topic x problem-type interaction - Unicode -> bare-LaTeX cleaner + audit + spot-check - Mirrors https://huggingface.co/datasets/blackhao0426/PutnamGAP --- CITATION.cff | 88 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 88 insertions(+) create mode 100644 CITATION.cff (limited to 'CITATION.cff') diff --git a/CITATION.cff b/CITATION.cff new file mode 100644 index 0000000..b77a346 --- /dev/null +++ b/CITATION.cff @@ -0,0 +1,88 @@ +cff-version: 1.2.0 +message: | + If you use the GAP framework or the PutnamGAP dataset, you MUST cite both + this work and the four MAA Putnam source books listed in the README under + "Citation". +title: "GAP — Generalization-and-Perturbation Framework for LLM Mathematical Reasoning Robustness" +authors: + - family-names: Hao + given-names: Yuren + - family-names: Wan + given-names: Xiang + - family-names: Zhai + given-names: ChengXiang +year: 2025 +url: "https://github.com/YurenHao0426/GAP" +repository-code: "https://github.com/YurenHao0426/GAP" +license: CC-BY-4.0 +preferred-citation: + type: article + title: | + An Investigation of Robustness of LLMs in Mathematical Reasoning: + Benchmarking with Mathematically-Equivalent Transformation of Advanced + Mathematical Problems + authors: + - family-names: Hao + given-names: Yuren + - family-names: Wan + given-names: Xiang + - family-names: Zhai + given-names: ChengXiang + year: 2025 + journal: "arXiv preprint arXiv:2508.08833" + url: "https://arxiv.org/abs/2508.08833" +references: + - type: book + title: "The William Lowell Putnam Mathematical Competition: Problems and Solutions 1938-1964" + authors: + - family-names: Gleason + given-names: A. M. + - family-names: Greenwood + given-names: R. E. + - family-names: Kelly + given-names: L. M. + year: 1980 + publisher: + name: Mathematical Association of America + notes: "MAA Problem Books, vol. 1; reprinted by AMS/MAA Press" + - type: book + title: "The William Lowell Putnam Mathematical Competition: Problems and Solutions 1965-1984" + authors: + - family-names: Alexanderson + given-names: Gerald L. + - family-names: Klosinski + given-names: Leonard F. + - family-names: Larson + given-names: Loren C. + year: 1985 + publisher: + name: Mathematical Association of America + notes: "MAA Problem Books, vol. 30" + - type: book + title: "The William Lowell Putnam Mathematical Competition 1985-2000: Problems, Solutions and Commentary" + authors: + - family-names: Kedlaya + given-names: Kiran S. + - family-names: Poonen + given-names: Bjorn + - family-names: Vakil + given-names: Ravi + year: 2002 + publisher: + name: Mathematical Association of America + notes: "MAA Problem Books, vol. 33" + - type: book + title: "The William Lowell Putnam Mathematical Competition 2001-2016: Problems, Solutions and Commentary" + authors: + - family-names: Kedlaya + given-names: Kiran S. + - family-names: Kane + given-names: Daniel M. + - family-names: Kane + given-names: Jonathan M. + - family-names: O'Dorney + given-names: Evan M. + year: 2020 + publisher: + name: American Mathematical Society (MAA Press) + notes: "MAA Problem Books, vol. 37" -- cgit v1.2.3