Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Autonomous Capability Assessment of Sequential Decision-Making Systems in Stochastic Settings
Pulkit Verma, Rushang Karia, Siddharth Srivastava
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 |
|---|---|---|---|
| Use QACE directed query synthesis instead of GLIB-G goal-sampling | learning probabilistic models of black-box SDM agents · Warehouse, Driving, First Responder, Elevator SDMAs | samples needed to reach GLIB-G plateau variational distance: at most 4% of the samples | Improved |
| Use QACE instead of GLIB (G and L) for learning time | model learning · Driving Agent, First Responder Robot, Elevator Control Agent | time to learn final model | Improved |
| Use QACE instead of GLIB on a simple SDMA with two capabilities, one deterministic | model learning · Warehouse Robot | learning time | Hurt |
| Use QACE instead of GLIB on SDMAs with many complex capabilities | model learning · First Responder Robot, Elevator Control Agent | variational distance to ground truth | Improved |
| Evaluate on test problems with more objects than training (<=7 objects) | QACE versus GLIB baselines · complex SDMA setups | variational distance | Improved |
| Use GLIB-L (lifted rules) instead of QACE | complex SDMAs · complex SDMAs | completion before timeout | Hurt |
| Add directed exploration using the partially learned model to build the state pool | QACE · state pool generation | sample complexity | Improved |
| Increase the hyperparameter eta (more collected triplets) | QACE probability estimation · effect probability learning | probability accuracy | Improved |
| Apply QACE to Cafe Server Robot with different vocabulary | QACE · Cafe Server Robot | variational distance | Improved |
Closest papers by meaning
- Agentick: A Unified Benchmark for General Sequential Decision-Making Agents (NeurIPS 2026 · similarity 0.84)
- Data Efficient Paradigms for Personalized Assessment of Black-Box Taskable AI Systems (AAAI 2024 · similarity 0.83)
- Generalization to New Sequential Decision Making Tasks with In-Context Learning (ICML 2024 · similarity 0.83)
- STL: Still Tricky Logic (for System Validation, Even When Showing Your Work) (NeurIPS 2024 · similarity 0.83)
- Hierarchical Planning and Learning for Robots in Stochastic Settings Using Zero-Shot Option Invention (AAAI 2024 · similarity 0.83)
- Foundations of Multi-Agent Learning in Dynamic Environments: Where Reinforcement Learning Meets Strategic Decision-Making (AAAI 2025 · similarity 0.82)
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.
