Formal Logic Deduction (FLD) is a project to enhance LLMs' reasoning capabilities via synthetically generated samples of logical reasoning, the most fundamental form of reasoning.
🎉 A new paper at NeurIPS 2024: Enhancing Reasoning Capabilities of LLMs via Principled Synthetic Logic Corpus.
💎 FLDx2 (Formal Logic Deduction Diverse), our most advanced corpus that substantially improves reasoning capabilities of state-of-the-art LLMs:
🎓 Built on well-grounded design principles, which integrate symbolic logic theory and previous empirical insights, resulting in diverse samples covering (i) multi-step deduction with unknown facts, (ii) diverse reasoning rules, (iii) diverse linguistic expressions, and (iv) challenging distractors.
🚀 Demonstrates substantial enhancement in LLM reasoning capabilities.
👊 Serves as a challenging benchmark asessing pure reasoning capabilities isolated from knowledge. Even GPT-4 can solve only about half of the problems.
For any reason where a GitHub pull request or an issue is not appropriate, feel free to email terufumi.morishita.wp[at]hitachi.com.
If you liked our project, please consider citing the following papers:
@inproceedings{morishita_2024_NeurIPS_FLD_diverse,
title={Enhancing Reasoning Capabilities of LLMs via Principled Synthetic Logic Corpus},
author={Terufumi Morishita and Gaku Morio and Atsuki Yamaguchi and Yasuhiro Sogawa},
booktitle={Annual Conference on Neural Information Processing Systems},
year={2024}
}
@inproceedings{morishita2024jfld,
title = {JFLD: A Japanese Benchmark for Deductive Reasoning based on Formal Logic},
author = {Morishita, Terufumi and Yamaguchi, Atsuki and Morio, Gaku and Hikaru, Tomonari and Osamu Imaichi and Sogawa, Yasuhiro},
booktitle = {Proceedings of the Joint International Conference on Computational Linguistics, Language Resources and Evaluation},
year = {2024}
}
@inproceedings{morishita2023fld,
title = {Learning Deductive Reasoning from Synthetic Corpus based on Formal Logic},
author = {Morishita, Terufumi and Morio, Gaku and Yamaguchi, Atsuki and Sogawa, Yasuhiro},
booktitle = {Proceedings of the 40th International Conference on Machine Learning},
year = {2023}
}
Formal Logic Deduction (FLD) is a project to enhance LLMs' reasoning capabilities via synthetically generated samples of logical reasoning, the most fundamental form of reasoning.
🎉 A new paper at NeurIPS 2024: Enhancing Reasoning Capabilities of LLMs via Principled Synthetic Logic Corpus.
💎 FLDx2 (Formal Logic Deduction Diverse), our most advanced corpus that substantially improves reasoning capabilities of state-of-the-art LLMs:
🎓 Built on well-grounded design principles, which integrate symbolic logic theory and previous empirical insights, resulting in diverse samples covering (i) multi-step deduction with unknown facts, (ii) diverse reasoning rules, (iii) diverse linguistic expressions, and (iv) challenging distractors.
🚀 Demonstrates substantial enhancement in LLM reasoning capabilities.
👊 Serves as a challenging benchmark asessing pure reasoning capabilities isolated from knowledge. Even GPT-4 can solve only about half of the problems.
For any reason where a GitHub pull request or an issue is not appropriate, feel free to email terufumi.morishita.wp[at]hitachi.com.
If you liked our project, please consider citing the following papers:
@inproceedings{morishita_2024_NeurIPS_FLD_diverse,
title={Enhancing Reasoning Capabilities of LLMs via Principled Synthetic Logic Corpus},
author={Terufumi Morishita and Gaku Morio and Atsuki Yamaguchi and Yasuhiro Sogawa},
booktitle={Annual Conference on Neural Information Processing Systems},
year={2024}
}
@inproceedings{morishita2024jfld,
title = {JFLD: A Japanese Benchmark for Deductive Reasoning based on Formal Logic},
author = {Morishita, Terufumi and Yamaguchi, Atsuki and Morio, Gaku and Hikaru, Tomonari and Osamu Imaichi and Sogawa, Yasuhiro},
booktitle = {Proceedings of the Joint International Conference on Computational Linguistics, Language Resources and Evaluation},
year = {2024}
}
@inproceedings{morishita2023fld,
title = {Learning Deductive Reasoning from Synthetic Corpus based on Formal Logic},
author = {Morishita, Terufumi and Morio, Gaku and Yamaguchi, Atsuki and Sogawa, Yasuhiro},
booktitle = {Proceedings of the 40th International Conference on Machine Learning},
year = {2023}
}