# Generating Effective CoT Traces for Mitigating Causal Hallucination

Source: [Generating Effective CoT Traces for Mitigating Causal Hallucination](<https://www.sota2.com/research/paper/69df1bbe7d0172140ec7bb0d>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-16T05:00:11.996Z

Authors: Yiheng Zhao, Jun Yan

Published: 2026-04-14T13:59:06.000Z

## Abstract

Although large language models (LLMs) excel in complex reasoning tasks, they suffer from severe causal hallucination in event causality identification (ECI), particularly in smaller models ($\\leq$1.5B parameters). A promising approach to address this issue is to fine-tune them with Chain-of-Thought (CoT) traces. However, there is currently a lack of CoT trace dataset available for ECI. In this paper, we first investigate the essential criteria that effective CoT traces should possess to mitigate causal hallucination in smaller models. We then design a pipeline to generate CoT traces that meet these criteria. Moreover, since there is currently no metric for quantifying causal hallucination, we also introduce a new metric, the Causal Hallucination Rate (CHR), to quantify causal hallucination, guide the formulation of effective CoT trace criteria, and validate the effectiveness of our pipeline. Our experiments show that fine-tuning with the CoT traces generated by our pipeline not only substantially reduces causal hallucination in smaller LLMs but also improves mean accuracy. Moreover, the fine-tuned models exhibit strong cross-dataset and cross-difficulty generalization, as well as robustness under misleading intervention prompts.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Event Causal Identification on EventStoryLine Document-level (inter-sentence event pairs)](<https://www.sota2.com/research/sota/event-causal-identification-on-eventstoryline-document-level-inter-sentence-event-pairs.md>): Event Causal Identification; EventStoryLine Document-level (inter-sentence event pairs); CHR: 93.94.

- [Event Causality Identification on Causal-TimeBank](<https://www.sota2.com/research/sota/event-causality-identification-on-causal-timebank.md>): Event causality identification; Causal-TimeBank; CHR: 84.55.

- [Event Causality Identification on MAVEN-ERE](<https://www.sota2.com/research/sota/event-causality-identification-on-maven-ere.md>): Event causality identification; MAVEN-ERE; CHR: 84.69.

- [Event Causal Identification on EventStoryLine](<https://www.sota2.com/research/sota/event-causal-identification-on-eventstoryline.md>): Event Causal Identification; EventStoryLine; CHR: 83.54.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.12748>)

[PDF](<https://arxiv.org/pdf/2604.12748.pdf>)