# Fine-Tuning without Performance Degradation

arXiv: https://arxiv.org/abs/2505.00913 · HTML: https://mlsanity.com/paper/2505.00913/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 29 of its 32 references that are in the map.
- First posted on arXiv 1 May 2025.

## Builds on (most cited first, 29 of 29)

- [Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor](https://mlsanity.com/paper/1801.01290/) (2018)
- [Conservative Q-Learning for Offline Reinforcement Learning](https://mlsanity.com/paper/2006.04779/) (2020)
- [D4RL: Datasets for Deep Data-Driven Reinforcement Learning](https://mlsanity.com/paper/2004.07219/) (2020)
- [Offline Reinforcement Learning with Implicit Q-Learning](https://mlsanity.com/paper/2110.06169/) (2021)
- [Soft Actor-Critic Algorithms and Applications](https://mlsanity.com/paper/1812.05905/) (2018)
- [AWAC: Accelerating Online Reinforcement Learning with Offline Datasets](https://mlsanity.com/paper/2006.09359/) (2020)
- [Efficient Online Reinforcement Learning with Offline Data](https://mlsanity.com/paper/2302.02948/) (2023)
- [Uncertainty-Based Offline Reinforcement Learning with Diversified Q-Ensemble](https://mlsanity.com/paper/2110.01548/) (2021)
- [Batch Policy Learning under Constraints](https://mlsanity.com/paper/1903.08738/) (2019)
- [Revisiting Few-sample BERT Fine-tuning](https://mlsanity.com/paper/2006.05987/) (2020)
- [Cal-QL: Calibrated Offline RL Pre-Training for Efficient Online Fine-Tuning](https://mlsanity.com/paper/2303.05479/) (2023)
- [Offline-to-Online Reinforcement Learning via Balanced Replay and Pessimistic Q-Ensemble](https://mlsanity.com/paper/2107.00591/) (2021)
- [To Tune or Not to Tune? Adapting Pretrained Representations to Diverse Tasks](https://mlsanity.com/paper/1903.05987/) (2019)
- [Hybrid RL: Using Both Offline and Online Data Can Make RL Efficient](https://mlsanity.com/paper/2210.06718/) (2022)
- [Better Fine-Tuning by Reducing Representational Collapse](https://mlsanity.com/paper/2008.03156/) (2020)
- [What Happens To BERT Embeddings During Fine-tuning?](https://mlsanity.com/paper/2004.14448/) (2020)
- [Jump-Start Reinforcement Learning](https://mlsanity.com/paper/2204.02372/) (2022)
- [Mildly Conservative Q-Learning for Offline Reinforcement Learning](https://mlsanity.com/paper/2206.04745/) (2022)
- [Policy Expansion for Bridging Offline-to-Online Reinforcement Learning](https://mlsanity.com/paper/2302.00935/) (2023)
- [Legged Robots that Keep on Learning: Fine-Tuning Locomotion Policies in the Real World](https://mlsanity.com/paper/2110.05457/) (2021)
- [Lifelong Generative Modeling](https://mlsanity.com/paper/1705.09847/) (2017)
- [A Closer Look at How Fine-tuning Changes BERT](https://mlsanity.com/paper/2106.14282/) (2021)
- [Unifying task specification in reinforcement learning](https://mlsanity.com/paper/1609.01995/) (2016)
- [Adaptive Behavior Cloning Regularization for Stable Offline-to-Online Reinforcement Learning](https://mlsanity.com/paper/2210.13846/) (2022)
- [The In-Sample Softmax for Offline Reinforcement Learning](https://mlsanity.com/paper/2302.14372/) (2023)
- [PROTO: Iterative Policy Regularized Offline-to-Online Reinforcement Learning](https://mlsanity.com/paper/2305.15669/) (2023)
- [Finetuning from Offline Reinforcement Learning: Challenges, Trade-offs and Practical Solutions](https://mlsanity.com/paper/2303.17396/) (2023)
- [Beyond Fine-Tuning: Transferring Behavior in Reinforcement Learning](https://mlsanity.com/paper/2102.13515/) (2021)
- [ENOTO: Improving Offline-to-Online Reinforcement Learning with Q-Ensembles](https://mlsanity.com/paper/2306.06871/) (2023)

## Abstract (arXiv)

Fine-tuning policies learned offline remains a major challenge in application domains. Monotonic performance improvement during fine-tuning is often challenging, as agents typically experience performance degradation at the early fine-tuning stage. The community has identified multiple difficulties in fine-tuning a learned network online, however, the majority of progress has focused on improving learning efficiency during fine-tuning. In practice, this comes at a serious cost during fine-tuning: initially, agent performance degrades as the agent explores and effectively overrides the policy learned offline. We show across a range of settings, many offline-to-online algorithms exhibit either (1) performance degradation or (2) slow learning (sometimes effectively no improvement) during fine-tuning. We introduce a new fine-tuning algorithm, based on an algorithm called Jump Start, that gradually allows more exploration based on online estimates of performance. Empirically, this approach achieves fast fine-tuning and significantly reduces performance degradations compared with existing algorithms designed to do the same.