Calibrated Stackelberg Games: Learning Optimal Commitments Against Calibrated Agents
Authors: Nika Haghtalab, Chara Podimata, Kunhe Yang
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | In our main technical result, we show that in CSGs, the principal can achieve utility that converges to the optimum Stackelberg value of the game both in finite and continuous settings, and that no higher utility is achievable. and Our work brings the perspective of calibrated forecasts to principal-agent games. We introduce Calibrated Stackelberg Games (CSG) a class that is more general than standard SGs and ask: Q1. What characterizes principal s optimal utility in CSGs? Q2. Are there natural forecasting algorithms for the agent that satisfy calibration? Our Contributions. We answer both questions completely. For Q1, we show that the principal s optimal utility converges exactly to V . For Q2, we give a general approach for obtaining a fine-grained any-time notion of calibration of independent interest and further specializing it to games. |
| Researcher Affiliation | Collaboration | Nika Haghtalab1, Chara Podimata2, and Kunhe Yang1 1University of California, Berkeley, {nika,kunheyang}@berkeley.edu 2MIT & Archimedes, podimata@mit.edu |
| Pseudocode | Yes | Algorithm 1: Explore-Then-Commit, Algorithm 2: Principal s Learning Algorithm for the Optimal Commitment, Algorithm 3: Lazy Gradient Descent without a Gradient (LAZYGDWOG), Algorithm 4: Approximate membership oracle for the conservative best response polytope (APPROXMEM), Algorithm 5: Post-Processing (POSTPROCESS) |
| Open Source Code | No | The paper does not contain any statement about making its source code publicly available or a link to a code repository. |
| Open Datasets | No | The paper is theoretical and does not involve empirical experiments with datasets, hence no information about public dataset access is provided. |
| Dataset Splits | No | The paper is theoretical and does not involve empirical experiments with datasets, hence no information about training/validation/test splits is provided. |
| Hardware Specification | No | The paper is theoretical and focuses on algorithm design and proofs; it does not describe empirical experiments that would require specific hardware specifications. |
| Software Dependencies | No | The paper mentions theoretical algorithms like ADANORMALHEDGE and LSV as conceptual tools, but it does not specify any software dependencies with version numbers for practical implementation or experimentation. |
| Experiment Setup | No | The paper is theoretical and focuses on algorithm design and proofs; it does not describe empirical experiments, therefore no experimental setup details like hyperparameters or training settings are provided. |