Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2502.09955 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910827361075200 |
|---|---|
| author | Drori, Iddo Longhitano, Gaston Mao, Mao Hyun, Seunghwan Zhang, Yuke Park, Sungjun Meeks, Zachary Zhang, Xin-Yu Segev, Ben Yong, Howard Verma, Nakul Shporer, Avi Amit, Alon Udell, Madeleine |
| author_facet | Drori, Iddo Longhitano, Gaston Mao, Mao Hyun, Seunghwan Zhang, Yuke Park, Sungjun Meeks, Zachary Zhang, Xin-Yu Segev, Ben Yong, Howard Verma, Nakul Shporer, Avi Amit, Alon Udell, Madeleine |
| contents | Reasoning LLMs such as OpenAI o1, o3 and DeepSeek R1 have made significant progress in mathematics and coding, yet find challenging advanced tasks such as International Mathematical Olympiad (IMO) combinatorics problems, Abstraction and Reasoning Corpus (ARC) puzzles, and Humanity's Last Exam (HLE) questions. We use a diverse inference approach that combines multiple models and methods at test time. We find that verifying mathematics and code problems, and rejection sampling on other problems is simple and effective. We automatically verify correctness of solutions to IMO problems by Lean, and ARC puzzles by code, and find that best-of-N effectively answers HLE questions. Our approach increases answer accuracy on IMO combinatorics problems from 33.3% to 77.8%, accuracy on HLE questions from 8% to 37%, and solves 80% of ARC puzzles that 948 humans could not and 26.5% of ARC puzzles that o3 high compute does not. Test-time simulations, reinforcement learning, and meta-learning with inference feedback improve generalization by adapting agent graph representations and varying prompts, code, and datasets. Our approach is reliable, robust, and scalable, and in the spirit of reproducible research, we will make it publicly available upon publication. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_09955 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Diverse Inference and Verification for Advanced Reasoning Drori, Iddo Longhitano, Gaston Mao, Mao Hyun, Seunghwan Zhang, Yuke Park, Sungjun Meeks, Zachary Zhang, Xin-Yu Segev, Ben Yong, Howard Verma, Nakul Shporer, Avi Amit, Alon Udell, Madeleine Artificial Intelligence Reasoning LLMs such as OpenAI o1, o3 and DeepSeek R1 have made significant progress in mathematics and coding, yet find challenging advanced tasks such as International Mathematical Olympiad (IMO) combinatorics problems, Abstraction and Reasoning Corpus (ARC) puzzles, and Humanity's Last Exam (HLE) questions. We use a diverse inference approach that combines multiple models and methods at test time. We find that verifying mathematics and code problems, and rejection sampling on other problems is simple and effective. We automatically verify correctness of solutions to IMO problems by Lean, and ARC puzzles by code, and find that best-of-N effectively answers HLE questions. Our approach increases answer accuracy on IMO combinatorics problems from 33.3% to 77.8%, accuracy on HLE questions from 8% to 37%, and solves 80% of ARC puzzles that 948 humans could not and 26.5% of ARC puzzles that o3 high compute does not. Test-time simulations, reinforcement learning, and meta-learning with inference feedback improve generalization by adapting agent graph representations and varying prompts, code, and datasets. Our approach is reliable, robust, and scalable, and in the spirit of reproducible research, we will make it publicly available upon publication. |
| title | Diverse Inference and Verification for Advanced Reasoning |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2502.09955 |