NOVA: Fundamental Limits of Knowledge Discovery Through AI
Quick Answer
This paper shows that The NOVA framework models AI's knowledge discovery process, identifying failure modes like contamination and forgetting.
Quick Take
The NOVA framework models AI's knowledge discovery process, identifying failure modes like contamination and forgetting. It quantifies the cumulative generation cost for distinct discoveries, revealing diminishing returns as exploration advances, emphasizing the importance of expert guidance near exploration barriers.
Key Points
- NOVA models the 'generate, verify, accumulate, retrain' loop as adaptive sampling.
- Identifies failure modes: contamination, forgetting, exploration, and acceptance failures.
- Cumulative generation cost for D discoveries scales as R_cum(D)=Θ(c_gen D^α).
- Good-Turing estimation is a local diagnostic, not a long-term discovery estimator.
- Expert input is crucial near barriers to autonomous exploration.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Can AI systems discover genuinely new knowledge through iterative self improvement, and if so, at what cost? We introduce the NOVA framework, which models the common ``generate, verify, accumulate, retrain'' loop as an adaptive sampling process over a knowledge space. We identify sufficient conditions under which accumulated genuine knowledge eventually covers a finite domain, and show how their violations produce distinct failure modes: contamination, forgetting, exploration failure, and acceptance failure. We then analyze imperfect verification and identify a contamination trap: as easy-to-find knowledge is exhausted, the model mass assigned to new valid artifacts shrinks, so even small false-positive rates can cause invalid artifacts to enter the knowledge base faster than genuine discoveries. We clarify that Good--Turing estimation is a local batch-diversity diagnostic, not an estimator of the historically undiscovered valid mass that governs long-term discovery. Under a separate tail-equivalence assumption relating the model's effective discovery distribution to a Zipf law with exponent $\alpha>1$, we prove that the cumulative generation cost required to obtain $D$ distinct genuine discoveries satisfies $R_{\mathrm{cum}}(D)=\Theta(c_{\mathrm{gen}}D^\alpha)$, where $c_{\mathrm{gen}}$ is the per-candidate generation cost. This scaling law quantifies asymptotic diminishing returns as the discovery frontier advances. Finally, we formalize human amplification through guidance, generation, and verification, explaining why expert input is most valuable near autonomous exploration barriers.
| Subjects: | Artificial Intelligence (cs.AI); Information Theory (cs.IT) |
| Cite as: | arXiv:2605.15219 [cs.AI] |
| (or arXiv:2605.15219v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2605.15219 arXiv-issued DOI via DataCite |
Submission history
From: Salman Avestimehr [view email]
[v1]
Tue, 12 May 2026 21:37:09 UTC (330 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →Procedural Memory Distillation: Online Reflection for Self-Improving Language Models
Procedural Memory Distillation (PMD) enhances reinforcement learning by converting cross-episode signals into reusable memory, improving Qwen3-8B and OLMo3-Instruct-7B models by 3.8-5.5% on SCIKNOWEVAL and 7.9-13.6% on . The co-evolution of policy and memory allows for more effective self-supervision, demonstrating significant performance gains when both components are active.