Search papers, labs, and topics across Lattice.
The paper introduces DIA-HARM, a benchmark to evaluate disinformation detection across 50 English dialects using the D3 corpus, which contains 195K dialectal disinformation samples generated via linguistically grounded transformations. Evaluating 16 models reveals that human-written dialectal content significantly degrades detection performance (1.4-3.6% F1), while AI-generated dialectal content is more robust; multilingual models like mDeBERTa exhibit strong cross-dialectal generalization, unlike monolingual models. This highlights the vulnerability of current disinformation detectors to dialectal variation.
Disinformation detectors trained on Standard American English can suffer catastrophic performance drops (>33% F1) when faced with other English dialects, potentially disadvantaging hundreds of millions of speakers.
Harmful content detectors-particularly disinformation classifiers-are predominantly developed and evaluated on Standard American English (SAE), leaving their robustness to dialectal variation unexplored. We present DIA-HARM, the first benchmark for evaluating disinformation detection robustness across 50 English dialects spanning U.S., British, African, Caribbean, and Asia-Pacific varieties. Using Multi-VALUE's linguistically grounded transformations, we introduce D3 (Dialectal Disinformation Detection), a corpus of 195K samples derived from established disinformation benchmarks. Our evaluation of 16 detection models reveals systematic vulnerabilities: human-written dialectal content degrades detection by 1.4-3.6% F1, while AI-generated content remains stable. Fine-tuned transformers substantially outperform zero-shot LLMs (96.6% vs. 78.3% best-case F1), with some models exhibiting catastrophic failures exceeding 33% degradation on mixed content. Cross-dialectal transfer analysis across 2,450 dialect pairs shows that multilingual models (mDeBERTa: 97.2% average F1) generalize effectively, while monolingual models like RoBERTa and XLM-RoBERTa fail on dialectal inputs. These findings demonstrate that current disinformation detectors may systematically disadvantage hundreds of millions of non-SAE speakers worldwide. We release the DIA-HARM framework, D3 corpus, and evaluation tools: https://github.com/jsl5710/dia-harm
