Joint Consistency: A Unified Test-Time Aggregation Framework via Energy Minimization
Yunzhen Yao, Hongye Wang, Yahong Wang, Michael C. Gastpar, Bo Jiang, Lie He · 2026
Read Joint Consistency: A Unified Test-Time Aggregation Framework via Energy Minimization by Yunzhen Yao, Hongye Wang, Yahong Wang, Michael C. Gastpar, Bo Jiang, Lie He (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-06219.txt.
Description
Yunzhen Yao and five co-authors unify test time aggregation of reasoning traces under one energy minimization frame. Models can sample many reasoning traces and combine them into a final answer, but most methods judge each trace alone or just count answer frequencies. That ignores how candidates compare with each other, information that could sharpen the final choice.
Their Joint Consistency method casts aggregation as a constrained Ising type energy problem. Independent evaluation signals act as external fields, while pairwise comparisons, gathered from a language model judge, act as interactions between candidates. Voting and weighted aggregation fall out as special cases of the frame. An approximation strategy keeps the interaction modeling practical at scale, and experiments on math and code benchmarks beat existing baselines.
Catalog record
| CREATOR | Yunzhen Yao, Hongye Wang, Yahong Wang, Michael C. Gastpar, Bo Jiang, Lie He |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.06219