Research map / Reinforcement learning / Policy gradient and actor-critic
Trajectory reuse in policy gradients
Reuse past trajectories and aggregate gradients for faster, more stable policy optimization.
68 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Policy gradient and actor-critic, keyword cluster “policy gradient · gradients · variance”
State History Tried Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
15 papers in 2023–24 and 16 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 4.4% to 3.5%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · Q-Prop: Sample-Efficient Policy Gradient with An Off-Policy Critic (ICLR 2017 · 117 citations)
- 2018 · Global Convergence of Policy Gradient Methods for the Linear Quadratic Regulator (ICML 2018 · 274 citations)
- 2019 · Understanding the Impact of Entropy on Policy Optimization (ICML 2019 · 44 citations)
- 2020 · Implementation Matters in Deep RL: A Case Study on PPO and TRPO (ICLR 2020 · 116 citations)
- 2021 · Phasic Policy Gradient (ICML 2021 · 51 citations)
- 2022 · A Parametric Class of Approximate Gradient Updates for Policy Optimization (ICML 2022 · 0 citations)
- 2023 · The Sufficiency of Off-Policyness and Soft Clipping: PPO Is Still Insufficient according to an Off-Policy Measure (AAAI 2023 · 20 citations)
- 2024 · Identifying Policy Gradient Subspaces (ICLR 2024 · 1 citation)
- 2025 · Convergence Analysis of Policy Gradient Methods with Dynamic Stochasticity (ICML 2025 · 0 citations)
- 2026 · Reparameterization Proximal Policy Optimization (ICML 2026 · 0 citations)
Earliest papers here
- Q-Prop: Sample-Efficient Policy Gradient with An Off-Policy Critic (ICLR 2017 · 117 citations)
- OFFER: Off-Environment Reinforcement Learning (AAAI 2017 · 26 citations)
- Interpolated Policy Gradient: Merging On-Policy and Off-Policy Gradient Estimation for Deep Reinforcement Learning (NeurIPS 2017 · 8 citations)
What has been tried
53 papers here report 503 measured changes (a component added, removed or swapped, compared with a baseline): helped 60%, no clear effect 10%, hurt 22%, mixed 8%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use a too-small regularization coefficient | CrossDocked2020, three protein targets · Vina score | Mixed | Stabilizing Policy Gradients for Stochastic Differential Equations via Consistency with Perturbation Process (ICML 2024 · 1 citation) | |
| Add observation normalization, penultimate normalization, and TD-error scaling | continuous-control benchmarks · episodic return | Helped | Deep Policy Gradient Methods Without Batch Updates, Target Networks, or Replay Buffers (NeurIPS 2024 · 1 citation) | |
| use schedule B with 1 iteration per phase instead of schedule A with 200 iterations per phase | Swimmer-v5 deterministic deployment (PB) · return | 61.14 ± 1.1 -> 60.40 ± 2.31 | Hurt | Convergence Analysis of Policy Gradient Methods with Dynamic Stochasticity (ICML 2025 · 0 citations) |
| Initialize the minimum cap using different policy-return estimates | Betting Game, CVaR α=0.2 · CVaR performance | Mixed | Return Capping: Sample Efficient CVaR Policy Gradient Optimisation (ICML 2025 · 0 citations) | |
| add entropy regularization with coefficient 5e-3 | Reorientation · successes | up to 16.5% | Helped | SAPG: Split and Aggregate Policy Gradients (ICML 2024 · 0 citations) |
| Remove stop-gradient on the policy input | Myriad · final performance | No clear effect | Do Transformer World Models Give Better Policy Gradients? (ICML 2024 · 0 citations) | |
| Replace MPO's tuned hyperparameters with WPO-matched hyperparameters | DeepMind Control Suite · performance | No clear effect | Wasserstein Policy Optimization (ICML 2025) |
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 31 papers; from OpenAlex and the papers' affiliations).
- University of Alberta · 3 papers
- Polytechnic Institute of Milan · 2 papers
- Politecnico di Milano · 2 papers
- University of California, San Diego · 2 papers
- Amazon · 2 papers
- Google DeepMind · 2 papers
- University of Maryland, College Park · 2 papers
Use it
Download BibTeX of 40 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Variance-reduced policy gradients (79 papers)
- Alternative policy update rules (50 papers)
- Policy gradients for special settings (45 papers)
Elsewhere in the topic
- Model-based RL with guarantees (245 papers)
- Regularized scalable deep value learning (216 papers)
- Temporal-difference and discounting theory (143 papers)
- Tree search and policy optimization (109 papers)
- Safety layers and shielding (85 papers)
- Actor-critic theory (76 papers)
- Off-policy actor-critic with expressive policies (71 papers)
- Distributional value learning (60 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).