# Beyond the Evidence Coherence Trap: Agent-Driven Evidence Selection for Temporally Disordered Fake News Detection

DevFeed: [Beyond the Evidence Coherence Trap: Agent-Driven Evidence Selection for Temporally Disordered Fake News Detection](<https://devfeed.tech/articles/beyond-the-evidence-coherence-trap-agent-driven-evidence-selection-for-temporally-disordered-fake-news-detection-103185.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.36764>)

Author: Chi Wang, Min Gao, Yifei Zhang, Bincheng Gu, Zongwei Wang, Wentao Li

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [ReAct agent Python](<https://devfeed.tech/topics/react-agent-python.md>), [AI-driven information retrieval](<https://devfeed.tech/topics/ai-driven-information-retrieval.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [coherence](<https://devfeed.tech/tags/coherence.md>), [consistency](<https://devfeed.tech/tags/consistency.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [detection](<https://devfeed.tech/tags/detection.md>), [fake-news](<https://devfeed.tech/tags/fake-news.md>), [inference](<https://devfeed.tech/tags/inference.md>), [llm](<https://devfeed.tech/tags/llm.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [narration](<https://devfeed.tech/tags/narration.md>), [relevance](<https://devfeed.tech/tags/relevance.md>), [reranker](<https://devfeed.tech/tags/reranker.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [selection](<https://devfeed.tech/tags/selection.md>), [temporal](<https://devfeed.tech/tags/temporal.md>), [training](<https://devfeed.tech/tags/training.md>), [verification](<https://devfeed.tech/tags/verification.md>)

## AI overview

Temporally disordered fake news can preserve the facts of individual events while rearranging their chronology. When similar events recur, retrieval may surface evidence from the wrong period, creating an "evidence coherence trap" that makes a false timeline seem locally credible. The authors introduce ChronoFake, a source-disjoint benchmark of 30,706 articles designed to distinguish temporal manipulation from valid narration, including flashbacks. Their DelSA method refines evidence selection during training and uses a frozen rulebook for one-pass selection at inference. An LLM characterizes candidate evidence and extracts the article's implied chronology; a verifier-trained reranker balances local relevance with timeline consistency, and a classifier compares article and evidence chronologies for global conflicts. The abstract reports that DelSA outperforms existing methods on ChronoFake and remains competitive on PolitiFact and GossipCop.

## Source excerpt

arXiv:2610.36764v1 Announce Type: new Abstract: Fake news undermines public trust and the integrity of online information ecosystems. Temporally disordered fake news is especially elusive because it rearranges factual events into a false chronology. Individual events may remain factually valid even when their temporal relations are misrepresented. When similar events involving the same entity recur, retrieval may return semantically plausible evidence from the wrong period. Consequently, independent event verification may rely on locally coherent but temporally mismatched evidence and incorrectly validate the article. We call this the \textit{evidence coherence trap}. Because existing benchmarks rarely isolate this setting, we construct ChronoFake, a source-disjoint benchmark of 30,706 articles derived from temporally anchored real events. It distinguishes temporal manipulation from valid narration, including flashbacks. To address this, we propose DelSA, a \underline{D}ocument S\underline{el}ection \underline{S}elf-Updating \underline{A}gent that iteratively refines evidence selection during training and applies a frozen rulebook for one-pass evidence selection at inference. DelSA uses an LLM to characterize candidate evidence, while a verifier-trained reranker balances local relevance against timeline consistency. Finally, the classifier compares the article-implied chronology extracted by the LLM with the evidence-derived chronology to reveal global temporal conflicts. DelSA outperforms existing methods on ChronoFake, while remaining competitive on PolitiFact and GossipCop.