DeepMind paper: model bidding auction hits 99.2% of optimal routing without knowing which model is best
rohanpaul_ai · x · 2026-10-23
A new Google DeepMind paper proposes letting AI models bid money for each prompt, with the highest bidder answering — achieving 99.2% of best-possible routing value when you don't know which model suits a question.
- Simulated auctions across 110 models on 30,000+ EmbedLLM prompts: full bidding reached 99.2% of optimal value vs 89.4% for routing each prompt type to its best model.
- On each method's worst prompt cluster, the auction kept 96.0% while the single best model kept just 6.0%.
Theoretically, the paper shows two features shrink the price-of-anarchy in agentic markets: a competitive floor (distilled cheaper models performing close to the best means multiple bidders value prompts highly, driving welfare loss toward zero) and near-independence of values conditioned on prompt difficulty, giving an efficiency guarantee of 1−1/e² of optimal welfare versus the general 1−1/e.
Practical takeaway: for model marketplaces, a simple pay-your-bid auction works near-optimally when the bidding pool has several strong models.
Related event: DeepMind paper: first-price auctions route AI queries at 99.2% of optimal(3 posts)→
More from Research
- Expert: this class of bio models needs drastically different evaluation schemes — anshulkundaje · 2026-10-23
- LSA: learnable spectral activations beat SIREN and FINER on INR reconstruction — orlitany · 2026-10-23
- New embedding approach loses on benchmarks but beats SAE-based ones qualitatively — mayfer · 2026-10-23
- VerifierBench: 222 Tasks Expose How Unreliable Agent-Generated Verifiers Are — Jianing Yin · 2026-10-23
- LM2LM: Frozen World Model Provides Scalable RL Rewards for AI Self-Improvement — Minhua Lin · 2026-10-23
- Stanford prof: LLM-fused models need radically new evals to separate memory from reasoning — anshulkundaje · 2026-10-23