analiza aiqbench
MaSCoD introduces a multi-agent framework for causal discovery that explicitly addresses the premature omission of relevant causal relations by organizing structural patterns before edge judgment. The study reveals that while supplying structural hypotheses improves recall and F1 scores, it also increases false-positive rates in a dataset- and model-dependent manner, challenging the notion of uniform superiority across LLM backbones. These findings offer critical insights for researchers aiming to balance omission control with precision in complex causal graph generation tasks.
MaSCoD: A Multi-Agent Framework for Structural-Context-Guided Candidate Causal Graph Generation
arXiv:2609.19944v1 Announce Type: new Abstract: Large language models (LLMs) have been applied to causal discovery, but candidate-graph generation rarely treats premature omission of potentially relevant causal relations as an explicit design objective. We propose MaSCoD, a multi-agent framework that organizes candidate third variables and local structural patterns before direct-edge judgment. We evaluate MaSCoD on Auto-MPG, DWD, and Sachs using GPT-5.4 as the primary backbone and GPT-4o for replication. MaSCoD exhibits a dataset- and backbone-dependent retention-selectivity profile rather than uniform superiority. Across all six dataset-backbone settings, Full, which supplies structural hypotheses before direct-edge judgment, achieved higher mean Recall and F1 than No Phase 1, which instead constructs them within the judgment procedure, while also increasing false-positive rates. Additional reference-edge retention over all evaluated baselines was observed on DWD with GPT-5.4 and on Sachs with GPT-4o, rather than uniformly across settings. Partial ablations showed that supplying both information components did not always outperform supplying only one. For GPT-5.4, stage-wise analysis showed that the Full-No Phase 1 retention gap was already present after direct-edge judgment, while reconciliation introduced additional reference-edge loss for Full on Sachs. These findings support structural pre-organization as an explicit design and evaluation target for omission control and motivate evaluating context construction jointly with its utilization in judgment.
Abstract:Large language models (LLMs) have been applied to causal discovery, but candidate-graph generation rarely treats premature omission of potentially relevant causal relations as an explicit design objective. We propose MaSCoD, a multi-agent framework that organizes candidate third variables and local structural patterns before direct-edge judgment. We evaluate MaSCoD on Auto-MPG, DWD, and Sachs using GPT-5.4 as the primary backbone and GPT-4o for replication. MaSCoD exhibits a dataset- and backbone-dependent retention-selectivity profile rather than uniform superiority. Across all six dataset-backbone settings, Full, which supplies structural hypotheses before direct-edge judgment, achieved higher mean Recall and F1 than No Phase 1, which instead constructs them within the judgment procedure, while also increasing false-positive rates. Additional reference-edge retention over all evaluated baselines was observed on DWD with GPT-5.4 and on Sachs with GPT-4o, rather than uniformly across settings. Partial ablations showed that supplying both information components did not always outperform supplying only one. For GPT-5.4, stage-wise analysis showed that the Full-No Phase 1 retention gap was already present after direct-edge judgment, while reconciliation introduced additional reference-edge loss for Full on Sachs. These findings support structural pre-organization as an explicit design and evaluation target for omission control and motivate evaluating context construction jointly with its utilization in judgment.
31 pages, 4 figures, 18 tables. The first two authors contributed equally
Artificial Intelligence (cs.AI)
arXiv:2609.19944 [cs.AI]
(or arXiv:2609.19944v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2609.19944 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.
Text integral sub licența deschisă a publicației · 309 cuvinte
Subiecte conexe
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · 13 oct. 2026
Evaluating AI Agents: A production blueprint with Strands and AgentCore
AWS · 23 iul. 2026
Links #5: 2026/07
LessWrong · 31 iul. 2026
Six Principles of Voice Design: Designing for the Ear and Encoding It into Agent Script
Salesforce Blog · 8 oct. 2026