Research map / Reinforcement learning / Policy gradient and actor-critic
Off-policy evaluation for slate recommendation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PI-OPT instead of SUP for pointwise ranking from page-level feedback | MSLR-WEB10K ranking · MSLR-WEB10K | NDCG@3 | Hurt |
| Use PI-OPT instead of SUP for pointwise ranking from page-level feedback | MSLR-WEB10K ranking · MSLR-WEB10K | ERR@3 | Helped |
| Use wPI rather than wIPS for slate off-policy evaluation | MSLR-WEB30K semi-synthetic ranking · MSLR-WEB30K | RMSE | Helped |
| Use wPI rather than DM when enough logged data is available | MSLR-WEB30K semi-synthetic ranking · MSLR-WEB30K | RMSE | Helped |
| Use wPI rather than DM at very small sample sizes | MSLR-WEB30K semi-synthetic ranking · MSLR-WEB30K | RMSE | Mixed |
| Apply PI despite violating the linearity assumption | Real-world search ranking evaluation · Time-to-success and UtilityRate | RMSE | Helped |
| Apply PI rather than DM at moderate sample sizes | Real-world search ranking evaluation · Time-to-success and UtilityRate | RMSE | Helped |
| Use wPI for ERR even though ERR violates linearity | MSLR-WEB30K semi-synthetic ranking · ERR | RMSE | Hurt |
Closest papers by meaning
- Distributional Off-Policy Evaluation for Slate Recommendations (AAAI 2024 · similarity 0.94)
- Control Variates for Slate Off-Policy Evaluation (NeurIPS 2021 · similarity 0.91)
- Diversified Recommendations for Agents with Adaptive Preferences (NeurIPS 2022 · similarity 0.88)
- Uncertainty-Aware Instance Reweighting for Off-Policy Learning (NeurIPS 2023 · similarity 0.87)
- Towards Off-Policy Learning for Ranking Policies with Logged Feedback (AAAI 2022 · similarity 0.87)
- Off-Policy Evaluation for Ranking Policies under Deterministic Logging Policies (ICLR 2026 · similarity 0.87)
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.