Sign in to view source links and access this dataset
Description
BenSyc is a culturally grounded benchmark for evaluating conversational sycophancy and social reinforcement in Bengali-context online conversations. The dataset preserves naturally occurring Bengali, Banglish, English, code-switching, slang, and emojis. It was authored by Sajib-006 and last updated on June 8, 2026.
Use Cases
Benchmarking conversational sycophancy in AI models based on the described social reinforcement phenomena.
Evaluating multilingual and code-switching language model performance based on the described mix of Bengali, Banglish, and English.
Studying culturally specific online communication patterns based on the described inclusion of slang and emojis.
Training models for human alignment in Bengali conversational contexts based on the benchmark's stated purpose.
Strengths
Focuses on a specific, culturally relevant NLP phenomenon: conversational sycophancy in Bengali contexts.
Preserves authentic linguistic features including code-switching, slang, and emojis as described.
Has a concept DOI (10.5281/zenodo.20392113) linking it to an academic publication.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
Source
huggingface
Freshness
Last updated 2026-06-08 18:14:39; freshness should be verified.
Geography
Likely focused on Bengali-speaking contexts.
License is unknown; users should verify usage rights.