Reasoning Consensus: Structural Ensembling of LLM Reasoning via Weighted DAG Aggregation
Amruta Parulekar, Jinu Lee, Dilek Hakkani-Tür, Hari Sundaram
cs.CL, cs.AI, cs.LG
Submitted: 2026-07-30
License: http://creativecommons.org/licenses/by-sa/4.0/
The gist: Large Language Models (LLMs) explore problems through chain-of-thought, but this exploration is buried in unstructured prose.
Terminology
Abstract
Large Language Models (LLMs) explore problems through chain-of-thought, but this exploration is buried in unstructured prose. On high-stakes tasks, users cannot tell which steps are well-supported, which alternatives were seriously considered, or how the final conclusion compares to those the model discarded. We propose a framework that ensembles the reasoning structure, not just the answers, of multiple LLMs by weighted merging of Directed Acyclic Graphs (DAGs) extracted from reasoning chains. We weight each step by how many traces independently attest to it, to return "Consensus Reasoning". Across six benchmarks spanning statutory interpretation, graduate-level science, narrative multi-hop reasoning, and first-order logic, our ensemble outperforms a matched-budget majority-vote baseline, with a maximum accuracy gain of 3.1% on MuSR-MM (narrative multi-hop reasoning). On a single model, the framework matches or exceeds self-consistency at the same trace budget while additionally exposing an inspectable consensus reasoning graph. Ensemble weights correlate with LLM-judge rankings of reasoning quality at Spearman rho = 0.30 - 0.51, and consensus subgraphs are preferred over alternatives leading to the majority-vote answer in 54.4-65.4% of head-to-head comparisons across five of six datasets. We observe that our framework can also be used to analyze diverse reasoning perspectives for a problem.
Sources
- ReConcile: Round-Table Conference Improves Reasoning via Consensus among Diverse LLMs
- The Llama 3 Herd of Models
- FOLIO: Natural Language Reasoning with First-Order Logic
- A Dataset for Statutory Reasoning in Tax Law Entailment and Question Answering
- Efficient Dynamic Ensembling for Multiple LLM Experts
- Mistral 7B
- ARGORA: Orchestrated Argumentation for Causally Grounded LLM Reasoning and Decision Making
- ReasoningFlow: Semantic Structure of Complex Reasoning Traces
- OpenAI o1 System Card
- Ensembling Large Language Models with Process Reward-Guided Tree Search for Better Complex Reasoning
- Qwen2.5 Technical Report
- Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
- Qwen3 Technical Report
- Multi-LLM Collaborative Search for Complex Problem Solving
- Beyond Accuracy: Evaluating Strategy Diversity in LLM Mathematical Reasoning
- Tree of Thoughts: Deliberate Problem Solving with Large Language Models
- GPQA: A Graduate-Level Google-Proof Q&A Benchmark
- HalluGuard: Demystifying Data-Driven and Reasoning-Driven Hallucinations in LLMs
- MuSR: Testing the Limits of Chain-of-thought with Multistep Soft Reasoning
- Gemma 3 Technical Report
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering