Lost in Interpretation: The Plausibility-Faithfulness Trade-off in Cross-Lingual Explanations
Quick Take
Cross-lingual explanations show a trade-off between plausibility and faithfulness in multilingual LLMs.
Key Points
- English explanations often lack causal grounding.
- Comprehensiveness can degrade by up to 5.7x.
- Auditing should be done in the input language.
📖 Reader Mode
~2 min readAbstract:LLMs deployed multilingually are often audited via English explanations for non-English inputs. We evaluate extractive explanations ''where the model identifies input token spans as evidence alongside a generated rationale'' and uncover a systematic trade-off: English-pivot explanations can achieve higher span agreement with human rationales while their evidence becomes less causally grounded in the model's prediction, as measured by both comprehensiveness and sufficiency. Across 3 tasks, 5~languages, and 2~multilingual LLM families, we find that English explanations frequently produce fluent but loosely anchored rationales, with comprehensiveness degrading by up to 5.7x relative to native-language conditions - even as task accuracy remains stable across settings. For socially nuanced classification, English pivots also fail to preserve pragmatic cues, reducing both faithfulness and span agreement. We recommend auditing explanations in the input language, reporting multi-faceted faithfulness metrics beyond lexical overlap, and treating English rationales as communication summaries rather than faithful decision traces.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2605.19274 [cs.CL] |
| (or arXiv:2605.19274v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.19274 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Somnath Banerjee [view email]
[v1]
Tue, 19 May 2026 02:44:48 UTC (165 KB)
— Originally published at arxiv.org
More from arXiv cs.CL
See more →Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?
The reliability of LLM judges for evaluating deep research agents is critically assessed using the REFLECT benchmark.