aiqbench analysis
This research introduces "Nash decoding," a novel framework that reformulates text generation as a game-theoretic problem to achieve significantly higher likelihood scores than standard autoregressive methods. By treating token positions as players in a Nash equilibrium, the approach improves performance on question-answering benchmarks without requiring model fine-tuning, though it does incur additional test-time computational costs.
Nash Equilibrium Text: A Game-Theoretic Decoding Framework for Text Generation
arXiv:2610.05817v1 Announce Type: new Abstract: Text revision has become an integral component of large language models. This paper formulates revision such that it admits a Nash equilibrium: Token positions are players, vocabulary items are actions, and each player's utility is the language model's log conditional probability. We motivate the revision by showing that Nash equilibria can have exponentially higher likelihood than autoregressive outputs as the sequence length grows. We further propose Nash decoding, an algorithm that reaches an $\varepsilon$-Nash equilibrium in $O(1/\varepsilon)$ time given access to the joint probability of tokens conditioned on a prompt. In practice, we run Nash decoding using conditional probability estimates from large language models and evaluate the resulting equilibria on question-answering benchmarks. On CLAPNQ, PubMedQA, and CoQA, Nash equilibria obtained from masked language models achieve higher F1 and ROUGE scores than autoregressive models up to $18\times$ larger, without any fine-tuning or retraining, at the cost of additional test-time computation.
Abstract:Text revision has become an integral component of large language models. This paper formulates revision such that it admits a Nash equilibrium: Token positions are players, vocabulary items are actions, and each player's utility is the language model's log conditional probability. We motivate the revision by showing that Nash equilibria can have exponentially higher likelihood than autoregressive outputs as the sequence length grows. We further propose Nash decoding, an algorithm that reaches an $\varepsilon$-Nash equilibrium in $O(1/\varepsilon)$ time given access to the joint probability of tokens conditioned on a prompt. In practice, we run Nash decoding using conditional probability estimates from large language models and evaluate the resulting equilibria on question-answering benchmarks. On CLAPNQ, PubMedQA, and CoQA, Nash equilibria obtained from masked language models achieve higher F1 and ROUGE scores than autoregressive models up to $18\times$ larger, without any fine-tuning or retraining, at the cost of additional test-time computation.
34 pages, 6 figures, 11 tables. Code: this https URL
Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
arXiv:2610.05817 [cs.AI]
(or arXiv:2610.05817v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.05817 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 257 words
Related coverage
Social Theory and Agent Foundations
LessWrong · Oct 8, 2026
Wasserstein Equilibrium Decoding for Reliable Medical Visual Question Answering
arXiv · Jun 16, 2026
Game-Theoretic Disempowerment: Loss of Control to Agencyless AI
LessWrong · Sep 27, 2026
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · Oct 13, 2026