ConGrs：用共识图合成语言模型多次采样的长文本回答

View PDF HTML (experimental)

Abstract:Language models can be sampled multiple times to access the distribution underlying their responses, but existing methods cannot efficiently synthesize rich epistemic signals across different long-form responses. We introduce Consensus Graphs (ConGrs), a flexible DAG-based data structure that represents shared information, as well as semantic variation in a set of sampled LM responses to the same prompt. We construct ConGrs using a light-weight lexical sequence alignment algorithm from bioinformatics, supplemented by the targeted usage of a secondary LM judge. Further, we design task-dependent decoding methods to synthesize a single, final response from our ConGr data structure. Our experiments show that synthesizing responses from ConGrs improves factual precision on two biography generation tasks by up to 31% over an average response and reduces reliance on LM judges by more than 80% compared to other methods. We also use ConGrs for three refusal-based tasks requiring abstention on unanswerable queries and find that abstention rate is increased by up to 56%. We apply our approach to the MATH and AIME reasoning tasks and find an improvement over self-verification and majority vote baselines by up to 6 points of accuracy. We show that ConGrs provide a flexible method for capturing variation in LM responses and using the epistemic signals provided by response variation to synthesize more effective responses.

Comments:	Published at COLM 2026
Subjects:	Computation and Language (cs.CL)
Cite as:	arXiv:2510.03527 [cs.CL]
	(or arXiv:2510.03527v2 [cs.CL] for this version)
	https://doi.org/10.48550/arXiv.2510.03527 arXiv-issued DOI via DataCite

Submission history

From: Sayan Ghosh [view email]
[v1] Fri, 3 Oct 2025 21:50:08 UTC (555 KB)
[v2] Thu, 15 Oct 2026 22:47:08 UTC (546 KB)