Reflect, Retry, Reward: Self-Improving LLMs via Reinforcement Learning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918040160960512 |
|---|---|
| author | Bensal, Shelly Jamil, Umar Bryant, Christopher Russak, Melisa Kamble, Kiran Mozolevskyi, Dmytro Ali, Muayad AlShikh, Waseem |
| author_facet | Bensal, Shelly Jamil, Umar Bryant, Christopher Russak, Melisa Kamble, Kiran Mozolevskyi, Dmytro Ali, Muayad AlShikh, Waseem |
| contents | We explore a method for improving the performance of large language models through self-reflection and reinforcement learning. By incentivizing the model to generate better self-reflections when it answers incorrectly, we demonstrate that a model's ability to solve complex, verifiable tasks can be enhanced even when generating synthetic data is infeasible and only binary feedback is available. Our framework operates in two stages: first, upon failing a given task, the model generates a self-reflective commentary analyzing its previous attempt; second, the model is given another attempt at the task with the self-reflection in context. If the subsequent attempt succeeds, the tokens generated during the self-reflection phase are rewarded. Our experimental results show substantial performance gains across a variety of model architectures, as high as 34.7% improvement at math equation writing and 18.1% improvement at function calling. Notably, smaller fine-tuned models (1.5 billion to 7 billion parameters) outperform models in the same family that are 10 times larger. Our novel paradigm is thus an exciting pathway to more useful and reliable language models that can self-improve on challenging tasks with limited external feedback. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_24726 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Reflect, Retry, Reward: Self-Improving LLMs via Reinforcement Learning Bensal, Shelly Jamil, Umar Bryant, Christopher Russak, Melisa Kamble, Kiran Mozolevskyi, Dmytro Ali, Muayad AlShikh, Waseem Computation and Language We explore a method for improving the performance of large language models through self-reflection and reinforcement learning. By incentivizing the model to generate better self-reflections when it answers incorrectly, we demonstrate that a model's ability to solve complex, verifiable tasks can be enhanced even when generating synthetic data is infeasible and only binary feedback is available. Our framework operates in two stages: first, upon failing a given task, the model generates a self-reflective commentary analyzing its previous attempt; second, the model is given another attempt at the task with the self-reflection in context. If the subsequent attempt succeeds, the tokens generated during the self-reflection phase are rewarded. Our experimental results show substantial performance gains across a variety of model architectures, as high as 34.7% improvement at math equation writing and 18.1% improvement at function calling. Notably, smaller fine-tuned models (1.5 billion to 7 billion parameters) outperform models in the same family that are 10 times larger. Our novel paradigm is thus an exciting pathway to more useful and reliable language models that can self-improve on challenging tasks with limited external feedback. |
| title | Reflect, Retry, Reward: Self-Improving LLMs via Reinforcement Learning |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2505.24726 |