Research map / Reinforcement learning / Imitation learning
Multi-Modal Inverse Constrained Reinforcement Learning from a Mixture of Demonstrations
Guanren Qiao, Guiliang Liu, Pascal Poupart, zhiqiang xu
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 contrastive estimation (InfoNCE) to policy update (MMICRL-LD -> MMICRL) (ablation) | MMICRL · Blocked Half-Cheetah agent z0 | feasible cumulative rewards: 4.32E+3 -> 6.12E+3 | Improved |
| add contrastive estimation to policy update (ablation) | MMICRL · Blocked Half-Cheetah | constraint violation rate: 33%±24% / 34%±24% -> 0%±0% / 0%±0% | Improved |
| add contrastive estimation to policy update (ablation) | MMICRL · Blocked Ant agent z0 | feasible cumulative rewards: 1.82E+4/ 2.21E+4 -> 2.13E+4 / 2.17E+4 | Improved |
| add contrastive estimation to policy update (ablation) | MMICRL · Blocked Ant | constraint violation rate: 0%±0% / 0%±0% -> 0%±0% / 0%±0% | No clear effect |
| add contrastive estimation to policy update (ablation) | MMICRL · Blocked Swimmer agent z0 | feasible cumulative rewards: 2.66E+2 -> 4.07E+2 | Improved |
| add contrastive estimation to policy update (ablation) | MMICRL · Blocked Swimmer | constraint violation rate: 71%±16% / 34%±23% -> 55%±23% / 28%±19% | Improved |
| add contrastive estimation to policy update (ablation) | MMICRL · Blocked Walker | feasible cumulative rewards: 9.02E+2 / 5.12E+2 -> 8.98E+2 / 1.53E+3 | Mixed |
| add contrastive estimation to policy update (ablation) | MMICRL · Blocked Walker | constraint violation rate: 52%±25% / 50%±25% -> 31%±22% / 25%±22% | Improved |
| replace trajectory-level flow density identifier with state-action discriminator (InfoGAIL-ICRL) | MMICRL · Blocked Half-Cheetah agent z0 | feasible cumulative rewards: 6.12E+3 -> 2.22E+2 | Hurt |
| remove agent identification (MEICRL, single unified constraint) | MMICRL · Blocked Half-Cheetah | constraint violation rate: 0%±0% / 0%±0% -> 100%±0% / 50%±25% | Hurt |
| remove agent identification (MEICRL) | MMICRL · Blocked Walker | constraint violation rate: 31%±22% / 25%±22% -> 100%±0% / 100%±0% | Hurt |
| remove agent identification (B2CL/MEICRL) | MMICRL · Gridworld | constraint recovery | Hurt |
Closest papers by meaning
- Robust Inverse Constrained Reinforcement Learning under Model Misspecification (ICML 2024 · similarity 0.92)
- Provably Efficient Exploration in Inverse Constrained Reinforcement Learning (ICML 2025 · similarity 0.91)
- Confidence Aware Inverse Constrained Reinforcement Learning (ICML 2024 · similarity 0.91)
- Learning Constraints from Offline Demonstrations via Superior Distribution Correction Estimation (ICML 2024 · similarity 0.90)
- Simplifying Constraint Inference with Inverse Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Recent Advancements in Inverse Reinforcement Learning (AAAI 2024 · similarity 0.88)
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.
