Hallucinations remain challenging for large language models (LLMs), especially in factual generation. Existing mitigation methods often rely on retrieval, supervision, representation editing, or output-level decoding, while paying limited attention to how factual support is internally coordinated across attention heads. In this work, we study factual decoding from a head-level coordination perspective and find that unreliable generations are often associated with fragmented and unstable use of factual anchors across heads. To improve such coordination, we propose Head Coordination Graphs (HCG), a plug-and-play inference-time framework that models attention heads as interacting nodes. HCG uses explicit factual anchors as observable support signals to construct sparse head graphs and applies risk-gated propagation to adaptively regulate cross-head influence without parameter updates, additional training, or external knowledge. Experiments across multiple factuality and hallucination benchmarks demonstrate broad improvements in factual consistency across open-source LLMs, while analyses and ablations demonstrate the importance of head-level coordination, sparse graph construction, and risk-gated propagation.

Document-level Event Causality Identification (DECI) aims to identify causal relations among multiple events within unstructured text. Existing methods predominantly rely on local semantic similarity for independent event-pair discrimination, thereby overlooking the influence of the overall narrative backbone in the propagation of causal dependencies and the role differentiation of events within multi-cause/multi-effect structures. Therefore, we propose a suggest-verify-revise approach for document-level Event Causality Identification with narrative consistency (SVRECI). In the suggest stage, we integrate multi-dimensional heuristic causal suggestions generated by an LLM with structural suggestions derived from hypergraph modeling to provide multi-source initial support for candidate event pairs. In the verify stage, we introduce a Topological Hawkes process to perform constrained verification of narrative propagation consistency among events. In the revise stage, we construct a dynamically evolving document-level causal graph and incorporate a structure-aware dual-level contrastive learning mechanism at both the event and event-pair levels, iteratively reducing noisy edges over multiple iterations. Experimental results on EventStoryLine and Causal-TimeBank datasets demonstrate that our approach outperforms previous methods.