Research map / Reinforcement learning / Policy gradient and actor-critic
TAAC: Temporally Abstract Actor-Critic for Continuous Control
Haonan Yu, Wei Xu, Haichao Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add TAAC's two-stage closed-loop act-or-repeat policy and compare-through evaluation | SAC actor-critic · All task categories | final n-score: 0.68[0.08] -> 0.96[0.03] | Improved |
| Add TAAC's two-stage closed-loop act-or-repeat policy and compare-through evaluation | SAC actor-critic · All task categories | n-AUC: 0.46[0.03] -> 0.72[0.03] | Improved |
| Add persistent exploration through closed-loop action repetition | SAC actor-critic (TAAC-1td) · All task categories | final n-score: 0.68[0.08] -> 0.92[0.05] | Improved |
| Add compare-through multi-step evaluation instead of one-step Bellman backup (ablation) | TAAC · All task categories | final n-score: 0.92[0.05] -> 0.96[0.03] | Improved |
| Add compare-through multi-step evaluation instead of one-step Bellman backup (ablation) | TAAC · All task categories | n-AUC: 0.65[0.04] -> 0.72[0.03] | Improved |
| Use one-step Bellman backup instead of compare-through (ablation) | TAAC · All task categories | final n-score: 0.96[0.03] -> 0.92[0.05] | Hurt |
| Always bootstrap with an N-step target without importance correction (ablation) | TAAC, N=10 · All task categories | final n-score: 0.93[0.05] -> 0.79[0.12] | Hurt |
| Always bootstrap with an N-step target without importance correction (ablation) | TAAC, N=10 · All task categories | n-AUC: 0.68[0.03] -> 0.55[0.08] | Hurt |
| Replace SAC-Hybrid's Retrace operator with compare-through (ablation) | SAC-Hybrid · All task categories | final n-score: 0.69[0.08] -> 0.74[0.11] | Improved |
| Replace SAC-Hybrid's Retrace operator with compare-through (ablation) | SAC-Hybrid · All task categories | n-AUC: 0.50[0.04] -> 0.50[0.06] | No clear effect |
| Use closed-loop action repetition with TAAC's post-sampling comparison instead of SAC-Hybrid's parallel switching decision | SAC actor-critic · Manipulation | final n-score: 0.49[0.07] -> 0.99[0.01] | Improved |
| Use closed-loop action repetition with TAAC's post-sampling comparison instead of SAC-Hybrid's parallel switching decision | SAC actor-critic · Terrain | final n-score: 0.79[0.03] -> 1.00[0.03] | Improved |
Later papers that compared with this method
1 later accepted paper reports TAAC as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Select before Act: Spatially Decoupled Action Repetition for Continuous Control (ICLR 2025 · better on 6 of 6 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Langevin Soft Actor-Critic: Efficient Exploration through Uncertainty-Driven Critic Learning (ICLR 2025 · similarity 0.88)
- On the Convergence of Continuous Single-timescale Actor-critic (ICML 2025 · similarity 0.88)
- Overcoming Slow Decision Frequencies in Continuous Control: Model-Based Sequence Reinforcement Learning for Model-Free Control (ICLR 2025 · similarity 0.87)
- ACE: Off-Policy Actor-Critic with Causality-Aware Entropy Regularization (ICML 2024 · similarity 0.87)
- Actor-Free Continuous Control via Structurally Maximizable Q-Functions (NeurIPS 2025 · similarity 0.87)
- ReLU to the Rescue: Improve Your On-Policy Actor-Critic with Positive Advantages (ICML 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
