VecCISC: Improving Confidence-Informed Self-Consistency with Reasoning Trace Clustering and Candidate Answer Selection
James PetulloSonny GeorgeDylan CashmanNianwen XueComputer Science Department, Brandeis University [email protected] · 2026
Read VecCISC: Improving Confidence-Informed Self-Consistency with Reasoning Trace Clustering and Candidate Answer Selection by James PetulloSonny GeorgeDylan CashmanNianwen XueComputer Science Department, Brandeis University [email protected] (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-2605-08070.txt.
Description
James Petullo, Sonny George, Dylan Cashman, and Nianwen Xue cut the cost of a popular reasoning trick. Self-consistency samples many answers from a language model and votes. A weighted version scores each answer's confidence with a second model call, which improves accuracy but adds heavy overhead for every candidate trace.
VecCISC instead measures semantic similarity between reasoning traces and filters out traces that repeat others, degenerate, or hallucinate. Fewer traces then need confidence scoring, so the weighted vote stays cheap. The framework is lightweight and adaptive, aiming to keep most of the accuracy gain of confidence-informed voting at a fraction of its cost.
Catalog record
| CREATOR | James PetulloSonny GeorgeDylan CashmanNianwen XueComputer Science Department, Brandeis University [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.08070