Papers by Yiheng Zhao
Generating Effective CoT Traces for Mitigating Causal Hallucination (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models suffer from severe causal hallucination in event causality identification (ECI) there is currently no metric for quantifying causal hallucinonation for small models . |
| Approach: | They propose to fine-tune large language models with Chain-of-Thought (CoT) traces to mitigate hallucination in smaller models by introducing a new metric, the Causal Hallucinations Rate, which quantifies hallucinosity. |
| Outcome: | The proposed pipeline reduces causal hallucination in smaller models and improves mean accuracy under intentionally misleading intervention prompts. |
Mitigating Causal Bias in LLMs via Potential Outcomes Framework and Actual Causality Theory (2026.findings-eacl)
Copied to clipboard
| Challenge: | Large Language Models exhibit significant causal hallucination, but evaluation of their document-level ECI performance is lacking. |
| Approach: | They propose to use Large Language Models to evaluate their document-level ECI performance . they propose a framework to mitigate the causal bias associated with using LLMs . |
| Outcome: | The proposed framework significantly reduces the causal bias associated with using LLMs on ECI while also achieving superior performance. |