On Blame Attribution for Accountable Multi-Agent Sequential Decision Making
Authors: Stelios Triantafyllou, Adish Singla, Goran Radanovic
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | we experimentally: a) validate the qualitative properties of the studied blame attribution methods, and b) analyze their robustness to uncertainty. |
| Researcher Affiliation | Academia | Stelios Triantafyllou MPI-SWS strianta@mpi-sws.org Adish Singla MPI-SWS adishs@mpi-sws.org Goran Radanovic MPI-SWS gradanovic@mpi-sws.org |
| Pseudocode | No | The paper describes mathematical formulations and discusses algorithms in paragraph text, but does not include structured pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | No | The paper does not provide a specific link to source code, nor does it explicitly state that the code for the described methodology is publicly available in supplementary material or via an anonymous link. It mentions that "The supplementary material provides more details on the experimental setup and implementation." but not code access. |
| Open Datasets | No | The paper describes simulated environments rather than using named public datasets with concrete access information. "To demonstrate the efficacy of the studied blame attribution methods, we consider two environments, Gridworld and Graph, depicted in Fig. 1 and Fig. 2. Both environments are adapted from [53] and modified to be multi-agent." |
| Dataset Splits | No | The paper describes training processes and parameters for the agents within simulated environments but does not provide specific numerical dataset splits (e.g., percentages or counts for training, validation, and test sets), citations to predefined splits, or detailed splitting methodologies. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper does not provide specific software dependency details with version numbers (e.g., library names like PyTorch or TensorFlow with their respective versions). |
| Experiment Setup | Yes | In RP er M experiments we set α = 0.4. In robustness experiments, we only consider uncertainty over the personal policy of A1, and we set α = 0.2 and α = 0.5. To model uncertainty, we consider maximum estimation error ϵmax, and to obtain uncertainty sets P(πb), we sample (uniformly at random) bπb i (s) such that 1 2 bπb i (s) πbi(s) 1 ϵmax. |