Research map / Large language models / Agents, code and math
Autobidding Auctions with LLM-Powered Creatives
Bingzhe Wang, Bowei Zhang, Changyuan Yu, Qi Qi
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 timescale separation (slow leader updates) (ablation) | TTSL-PPO · AntM2C platform (GPT-4o surrogate) | revenue change (%): +2.14 -> +2.66 | Improved |
| add inferred dual variables to platform state (ablation) | TTSL-PPO · AntM2C platform (GPT-4o surrogate) | revenue change (%): +2.16 -> +2.66 | Improved |
| use GAE instead of Monte-Carlo returns (ablation) | TTSL-PPO · AntM2C platform (GPT-4o surrogate) | revenue change (%): +2.17 -> +2.66 | Improved |
| add timescale separation (ablation) | TTSL-PPO · End-to-end GPT-4o API | revenue change (%): +1.66 -> +2.04 | Improved |
| use TTSL-PPO instead of myopic greedy investment | GSP auction · AntM2C platform (GPT-4o surrogate) | revenue change (%): -7.09 -> +2.66 | Improved |
| use TTSL-PPO instead of uniform investment | GSP auction · AntM2C platform (GPT-4o surrogate) | revenue change (%): -9.17 -> +2.66 | Improved |
| invest uniformly in LLM enhancement vs no investment | GSP auction · AntM2C platform (GPT-4o surrogate) | social welfare change (%): +0.00 -> +4.12 | Mixed |
| use ODDB-R instead of Pacing | autobidding agents · Agent bidding subgame (10 agents, GSP, T=2000) | social welfare vs Truthful (%): -30.1% -> -29.3% | Improved |
| use ODDB-R instead of Pacing | autobidding agents · Agent bidding subgame (10 agents, GSP, T=2000) | convergence round: 1063 -> 841 | Improved |
| use stronger LLM enhancement (low to high capability) | TTSL-PPO · AntM2C platform | revenue change (%): +2.62 -> +2.70 | Improved |
| scale to 100 heterogeneous agents and 3 slots | TTSL-PPO · Scalability (100 agents, 3 slots) | revenue change (%): +2.66 -> +1.07 | Mixed |
| use TTSL-PPO instead of GFP no investment | GFP auction · AntM2C platform (GPT-4o surrogate) | revenue change (%): -0.32 -> +2.66 | Improved |
Closest papers by meaning
- LLM-Auction: Generative Auction towards LLM-Native Advertising (NeurIPS 2026 · similarity 0.90)
- AuctionNet: A Novel Benchmark for Decision-Making in Large-Scale Games (NeurIPS 2024 · similarity 0.88)
- Ad Auctions for LLMs via Retrieval Augmented Generation (NeurIPS 2024 · similarity 0.86)
- Increasing Revenue in Efficient Combinatorial Auctions by Learning to Generate Artificial Competition (AAAI 2025 · similarity 0.86)
- Auctionformer: A Unified Deep Learning Algorithm for Solving Equilibrium Strategies in Auction Games (ICML 2024 · similarity 0.85)
- Truthful Aggregation of LLMs with an Application to Online Advertising (NeurIPS 2025 · similarity 0.85)
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.
